diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/resolved_config.yaml b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/resolved_config.yaml index afb585887edaa9058a96f488ddfd87e69e4c4978..5a606074dce532c00fa05acd88d0c24752183119 100644 --- a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/resolved_config.yaml +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/resolved_config.yaml @@ -17,14 +17,13 @@ model: rkd_loss_weight: 0.5 rkd_loss_weight_schedule: null phases: -- name: train_5shot_seed0_processingline_12k - max_steps: 12000 +- name: train_5shot_seed0_processingline_5k + max_steps: 5000 save_steps: 0 save_step_list: - - 2000 + - 1000 + - 3000 - 5000 - - 8000 - - 12000 trainable: preset: processing_line_only tune_llm: false diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/seed0.yaml b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/seed0.yaml index b4861c0e642baa4a733e1e9d745c1efcea527142..41a0ec1a59538b9de348582c7370101e4b984765 100644 --- a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/seed0.yaml +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/seed0.yaml @@ -17,14 +17,13 @@ model: rkd_loss_weight: 0.5 rkd_loss_weight_schedule: null phases: -- name: train_5shot_seed0_processingline_12k - max_steps: 12000 +- name: train_5shot_seed0_processingline_5k + max_steps: 5000 save_steps: 0 save_step_list: - - 2000 + - 1000 + - 3000 - 5000 - - 8000 - - 12000 trainable: preset: processing_line_only tune_llm: false diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a9ea477b4c6d05620a765600fbaacbfa2c8c8951 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c9eb7c3011cbf3e7c8d2c438ce9a8c8014620393f03490c5df5166bf6e48bf7 +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/rng_state.pth b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..7ce1bc1f4794acacd86118a4425e4680f5a99b69 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e07d71857af8855ca9953510adf0990edf4d1e42bf678742556f1530ae39d259 +size 14645 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/scheduler.pt b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..cddb85352b4d053450624108786800cbad81d6cb --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e19ead5779e3f8f2f666b7793de3a03d118f57f1b16a97eeb87bd732d18b445 +size 1465 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a3758b2dac9cab1274647a0a3e4647bb197ec7fc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6f702e66e76c9c2a5773af5ec6b03a2aabb911ab03f04ce0b6ef1d2a8dca66a +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/rng_state.pth b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..7b88bcf5b04146bda01f0c86dc0b5084d0d6b6e5 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:aa67eb0b846a8b110515825320d8417c3b8601df002d53f4c043f61a6b3bef11 +size 14645 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/scheduler.pt b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..de4e41c4a6949df056e028b10a18e7f8cd7b7580 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:42d8e6d8e53e6b837084d0cc18fdd63ca9e40bd44a65d38eb52695647a8e84e8 +size 1465 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a3758b2dac9cab1274647a0a3e4647bb197ec7fc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6f702e66e76c9c2a5773af5ec6b03a2aabb911ab03f04ce0b6ef1d2a8dca66a +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..15465b8cc3eabe1f42caab01a97d6b12e58a8f43 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a744ff059a47ce006479e017d86e92e45454084ab8a994491f962aee2d700ce +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/rng_state.pth b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..028a177f40344779e43d4da93a8e49bb81c37e0d --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ee05483c1b77b6ee84eeb1bd758810c3b2f3502216b646cea257d7f1de86be7 +size 14645 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/scheduler.pt b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..8cb18c42a19c524a920dd1ec232118cfa1c2aa5a --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_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-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/config.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..6f73f8489294c58acc82d1351a17e5e3ff316e18 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5886094736594543, + -2.1559664577316586, + 0.7003580854548973 + ], + "std": [ + 1.6717149343420803, + 1.3743695399555862, + 0.000940069542359506 + ], + "q01": [ + -0.7517815060249613, + -5.497849509089705, + 0.6999776914424374 + ], + "q99": [ + 6.060945690807028, + -0.5331499932493624, + 0.7028864260472257 + ] + }, + "base_rotation": { + "max": [ + 0.0, + 0.0, + 1.0, + 1.0 + ], + "min": [ + 0.0, + 0.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0, + 0.0, + 0.24488939171811902, + 0.5968549204077488 + ], + "std": [ + 0.0, + 0.0, + 0.6659970734507764, + 0.37443298392262264 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 4.149924358336069e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.25087486209693133, + -0.005738920462036393, + 0.43643948866361787 + ], + "std": [ + 0.15741544532518012, + 0.2547603106793105, + 0.2050434492495558 + ], + "q01": [ + -0.013990057288082561, + -0.4824599092623036, + 0.18825220353500022 + ], + "q99": [ + 0.5411989642841469, + 0.4754923657245658, + 0.6912323747882034 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22290331318740714, + 0.04579037037503981, + -0.05804094906527823, + 0.12347219024917105 + ], + "std": [ + 0.8519267931761344, + 0.2949394732270141, + 0.30026078832556913, + 0.16333086331829588 + ], + "q01": [ + -0.9957959450418518, + -0.5678869493126506, + -0.5318135475761875, + 0.0014160465686576505 + ], + "q99": [ + 0.9900692296898317, + 0.6357358195673684, + 0.5009847868015971, + 0.4287044482210845 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03203268930378034, + -0.032331639271968075 + ], + "std": [ + 0.011607049173153697, + 0.011306286745872529 + ], + "q01": [ + 0.006429109393656672, + -0.04056586943958638 + ], + "q99": [ + 0.040377860142451545, + -0.006593889698673258 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006050336852760097, + 0.002728916466628117, + -0.0017236811393598838, + 0.0 + ], + "std": [ + 0.11032167711449951, + 0.09824140301615687, + 0.07825208047515658, + 0.0 + ], + "q01": [ + -0.154721262094005, + -0.18955302988697728, + -0.14454065867754942, + 0.0 + ], + "q99": [ + 0.24492955247744197, + 0.2496797822190429, + 0.11966742325336034, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9294448715752568 + ], + "std": [ + 0.36897903503358326 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.21627562513311385 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.00826780165926883, + 0.003161997359379232, + -0.08157453651841007 + ], + "std": [ + 0.4545430944464769, + 0.41040151434986594, + 0.4102644251536544 + ], + "q01": [ + -0.9653185255999107, + -0.9255635199318754, + -0.8208437313341718 + ], + "q99": [ + 0.7987781084592441, + 0.9260844232330088, + 0.8357421812622201 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.007367024641295137, + -0.02060756343728361, + -0.001884833075998489 + ], + "std": [ + 0.1116768956941076, + 0.13249042272165212, + 0.12338908364725322 + ], + "q01": [ + -0.26894658663246845, + -0.39865607740339326, + -0.34393735191302655 + ], + "q99": [ + 0.3191674780171587, + 0.31133630397431267, + 0.33968476423051097 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3098994894722959 + ], + "std": [ + 0.9507718287498876 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7545620899936146 + ] + } + } + }, + "modalities": { + "video": { + "robot0_eye_in_hand": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_left": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_right": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + } + }, + "state": { + "base_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "base_rotation": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "end_effector_position_relative": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation_relative": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "gripper_qpos": { + "absolute": true, + "rotation_type": null, + "shape": [ + 2 + ], + "continuous": true + } + }, + "action": { + "base_motion": { + "absolute": true, + "rotation_type": null, + "shape": [ + 4 + ], + "continuous": true + }, + "control_mode": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + }, + "end_effector_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation": { + "absolute": true, + "rotation_type": "axis_angle", + "shape": [ + 3 + ], + "continuous": true + }, + "gripper_close": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + } + } + }, + "embodiment_tag": "new_embodiment" + } +} \ No newline at end of file diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a3758b2dac9cab1274647a0a3e4647bb197ec7fc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6f702e66e76c9c2a5773af5ec6b03a2aabb911ab03f04ce0b6ef1d2a8dca66a +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..f464a5a628e71ff8a8b3c93016f7e9a0ca875727 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_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": 0.2056944653391838, + "grad_norm": 0.9650156497955322, + "learning_rate": 1.08e-06, + "loss": 0.3957, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13596224263310433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10807111635804176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.37999560236930846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10, + "weighted_flow_matching_loss": 0.2056944653391838, + "weighted_rkd_loss": 0.18999780118465423 + }, + { + "epoch": 0.03355704697986577, + "flow_matching_loss": 0.24405399858951568, + "grad_norm": 1.1819026470184326, + "learning_rate": 2.28e-06, + "loss": 0.4373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13781716972589492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11078641712665557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3864207535982132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20, + "weighted_flow_matching_loss": 0.24405399858951568, + "weighted_rkd_loss": 0.1932103767991066 + }, + { + "epoch": 0.050335570469798654, + "flow_matching_loss": 0.18607925474643708, + "grad_norm": 1.1378169059753418, + "learning_rate": 3.48e-06, + "loss": 0.3642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12805165946483613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10023756325244904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3563408851623535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 30, + "weighted_flow_matching_loss": 0.18607925474643708, + "weighted_rkd_loss": 0.17817044258117676 + }, + { + "epoch": 0.06711409395973154, + "flow_matching_loss": 0.1954496458172798, + "grad_norm": 0.9604244232177734, + "learning_rate": 4.68e-06, + "loss": 0.3626, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12190236821770668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09053741469979286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33434214890003205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 40, + "weighted_flow_matching_loss": 0.1954496458172798, + "weighted_rkd_loss": 0.16717107445001603 + }, + { + "epoch": 0.08389261744966443, + "flow_matching_loss": 0.18862971812486648, + "grad_norm": 1.490982174873352, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12135515362024307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09131846278905868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33402876555919647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 50, + "weighted_flow_matching_loss": 0.18862971812486648, + "weighted_rkd_loss": 0.16701438277959824 + }, + { + "epoch": 0.10067114093959731, + "flow_matching_loss": 0.1686223641037941, + "grad_norm": 0.6995788812637329, + "learning_rate": 7.08e-06, + "loss": 0.33, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11892070844769478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08481515794992447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.322656574845314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 60, + "weighted_flow_matching_loss": 0.1686223641037941, + "weighted_rkd_loss": 0.161328287422657 + }, + { + "epoch": 0.1174496644295302, + "flow_matching_loss": 0.1650075815618038, + "grad_norm": 0.7736762762069702, + "learning_rate": 8.28e-06, + "loss": 0.3138, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10964272096753121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07826093062758446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2975463718175888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 70, + "weighted_flow_matching_loss": 0.1650075815618038, + "weighted_rkd_loss": 0.1487731859087944 + }, + { + "epoch": 0.1342281879194631, + "flow_matching_loss": 0.14050063565373422, + "grad_norm": 0.7896946668624878, + "learning_rate": 9.48e-06, + "loss": 0.2771, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10285875871777535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06756566911935806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.27328317910432814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 80, + "weighted_flow_matching_loss": 0.14050063565373422, + "weighted_rkd_loss": 0.13664158955216407 + }, + { + "epoch": 0.15100671140939598, + "flow_matching_loss": 0.12970626056194307, + "grad_norm": 0.6709607839584351, + "learning_rate": 1.068e-05, + "loss": 0.2633, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09869295060634613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06970738619565964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.267093288898468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 90, + "weighted_flow_matching_loss": 0.12970626056194307, + "weighted_rkd_loss": 0.133546644449234 + }, + { + "epoch": 0.16778523489932887, + "flow_matching_loss": 0.11140652000904083, + "grad_norm": 0.6956789493560791, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.2261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0861569106578827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05709078311920166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22940460294485093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 100, + "weighted_flow_matching_loss": 0.11140652000904083, + "weighted_rkd_loss": 0.11470230147242547 + }, + { + "epoch": 0.18456375838926176, + "flow_matching_loss": 0.09781593903899193, + "grad_norm": 0.7636014223098755, + "learning_rate": 1.308e-05, + "loss": 0.2177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09013360589742661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.059417485073208806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23968469649553298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 110, + "weighted_flow_matching_loss": 0.09781593903899193, + "weighted_rkd_loss": 0.11984234824776649 + }, + { + "epoch": 0.20134228187919462, + "flow_matching_loss": 0.08636336326599121, + "grad_norm": 0.6833075881004333, + "learning_rate": 1.428e-05, + "loss": 0.1919, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0791482873260975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.052746016159653665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2110425904393196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 120, + "weighted_flow_matching_loss": 0.08636336326599121, + "weighted_rkd_loss": 0.1055212952196598 + }, + { + "epoch": 0.2181208053691275, + "flow_matching_loss": 0.08181707337498664, + "grad_norm": 0.5722662210464478, + "learning_rate": 1.548e-05, + "loss": 0.1877, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0808029130101204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05019080862402916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2117966368794441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 130, + "weighted_flow_matching_loss": 0.08181707337498664, + "weighted_rkd_loss": 0.10589831843972206 + }, + { + "epoch": 0.2348993288590604, + "flow_matching_loss": 0.08008071929216384, + "grad_norm": 0.5205653309822083, + "learning_rate": 1.6680000000000003e-05, + "loss": 0.1788, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07563451901078225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.046241700649261475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19751073718070983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 140, + "weighted_flow_matching_loss": 0.08008071929216384, + "weighted_rkd_loss": 0.09875536859035491 + }, + { + "epoch": 0.2516778523489933, + "flow_matching_loss": 0.07708811983466149, + "grad_norm": 0.6222124099731445, + "learning_rate": 1.7879999999999998e-05, + "loss": 0.1711, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07161490805447102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04470515791326761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18793497681617738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 150, + "weighted_flow_matching_loss": 0.07708811983466149, + "weighted_rkd_loss": 0.09396748840808869 + }, + { + "epoch": 0.2684563758389262, + "flow_matching_loss": 0.07064299248158931, + "grad_norm": 0.5130728483200073, + "learning_rate": 1.908e-05, + "loss": 0.1537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06377769745886326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.038518281653523445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.166073676943779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 160, + "weighted_flow_matching_loss": 0.07064299248158931, + "weighted_rkd_loss": 0.0830368384718895 + }, + { + "epoch": 0.28523489932885904, + "flow_matching_loss": 0.0708860356360674, + "grad_norm": 0.44459694623947144, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06139358542859554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03676006477326155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15954723656177522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 170, + "weighted_flow_matching_loss": 0.0708860356360674, + "weighted_rkd_loss": 0.07977361828088761 + }, + { + "epoch": 0.30201342281879195, + "flow_matching_loss": 0.06621370911598205, + "grad_norm": 0.43706104159355164, + "learning_rate": 2.148e-05, + "loss": 0.1474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.062162948027253154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03814310748130083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16246900260448455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 180, + "weighted_flow_matching_loss": 0.06621370911598205, + "weighted_rkd_loss": 0.08123450130224227 + }, + { + "epoch": 0.3187919463087248, + "flow_matching_loss": 0.06586643978953362, + "grad_norm": 0.48638027906417847, + "learning_rate": 2.268e-05, + "loss": 0.1419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.058573270216584206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03490185085684061, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15204839110374452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 190, + "weighted_flow_matching_loss": 0.06586643978953362, + "weighted_rkd_loss": 0.07602419555187226 + }, + { + "epoch": 0.33557046979865773, + "flow_matching_loss": 0.06912102214992047, + "grad_norm": 0.45584234595298767, + "learning_rate": 2.3880000000000002e-05, + "loss": 0.1418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0558802418410778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03365606628358364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1454165495932102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 200, + "weighted_flow_matching_loss": 0.06912102214992047, + "weighted_rkd_loss": 0.0727082747966051 + }, + { + "epoch": 0.3523489932885906, + "flow_matching_loss": 0.06224445998668671, + "grad_norm": 0.516363799571991, + "learning_rate": 2.508e-05, + "loss": 0.1261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04999670423567295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027720438688993453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12771384567022323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 210, + "weighted_flow_matching_loss": 0.06224445998668671, + "weighted_rkd_loss": 0.06385692283511162 + }, + { + "epoch": 0.3691275167785235, + "flow_matching_loss": 0.06367993578314782, + "grad_norm": 0.4426209032535553, + "learning_rate": 2.628e-05, + "loss": 0.1292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.050844109430909154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02938362695276737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1310718484222889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 220, + "weighted_flow_matching_loss": 0.06367993578314782, + "weighted_rkd_loss": 0.06553592421114444 + }, + { + "epoch": 0.3859060402684564, + "flow_matching_loss": 0.061497534066438674, + "grad_norm": 0.4488156735897064, + "learning_rate": 2.748e-05, + "loss": 0.1214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04617831408977509, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02737251725047827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11972914338111877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 230, + "weighted_flow_matching_loss": 0.061497534066438674, + "weighted_rkd_loss": 0.059864571690559386 + }, + { + "epoch": 0.40268456375838924, + "flow_matching_loss": 0.05670501254498959, + "grad_norm": 0.4305253028869629, + "learning_rate": 2.868e-05, + "loss": 0.1146, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04511659294366836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02554646022617817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11577964574098587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 240, + "weighted_flow_matching_loss": 0.05670501254498959, + "weighted_rkd_loss": 0.057889822870492935 + }, + { + "epoch": 0.41946308724832215, + "flow_matching_loss": 0.052361448854207994, + "grad_norm": 0.39369747042655945, + "learning_rate": 2.9880000000000002e-05, + "loss": 0.1061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04291637688875198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021558509021997452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10739126279950142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 250, + "weighted_flow_matching_loss": 0.052361448854207994, + "weighted_rkd_loss": 0.05369563139975071 + }, + { + "epoch": 0.436241610738255, + "flow_matching_loss": 0.05451476089656353, + "grad_norm": 0.5052051544189453, + "learning_rate": 2.9999734259635676e-05, + "loss": 0.108, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04195186048746109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02312448900192976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10702821090817452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 260, + "weighted_flow_matching_loss": 0.05451476089656353, + "weighted_rkd_loss": 0.05351410545408726 + }, + { + "epoch": 0.45302013422818793, + "flow_matching_loss": 0.060980511084198954, + "grad_norm": 0.4307898283004761, + "learning_rate": 2.9998815663057245e-05, + "loss": 0.1147, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04224815517663956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023023180663585663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10751949027180671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 270, + "weighted_flow_matching_loss": 0.060980511084198954, + "weighted_rkd_loss": 0.053759745135903356 + }, + { + "epoch": 0.4697986577181208, + "flow_matching_loss": 0.050894477590918544, + "grad_norm": 0.395856112241745, + "learning_rate": 2.999724096969228e-05, + "loss": 0.1016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.039661315828561784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022144765593111515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10146739631891251, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 280, + "weighted_flow_matching_loss": 0.050894477590918544, + "weighted_rkd_loss": 0.050733698159456254 + }, + { + "epoch": 0.4865771812080537, + "flow_matching_loss": 0.05287523865699768, + "grad_norm": 0.3969660699367523, + "learning_rate": 2.9995010248422984e-05, + "loss": 0.1005, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03771056812256575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019884974136948584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09530610963702202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 290, + "weighted_flow_matching_loss": 0.05287523865699768, + "weighted_rkd_loss": 0.04765305481851101 + }, + { + "epoch": 0.5033557046979866, + "flow_matching_loss": 0.0479743555188179, + "grad_norm": 0.36045897006988525, + "learning_rate": 2.999212359682832e-05, + "loss": 0.0888, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03238320779055357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016917126439511775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08168354257941246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 300, + "weighted_flow_matching_loss": 0.0479743555188179, + "weighted_rkd_loss": 0.04084177128970623 + }, + { + "epoch": 0.5201342281879194, + "flow_matching_loss": 0.04859849065542221, + "grad_norm": 0.39492926001548767, + "learning_rate": 2.9988581141179787e-05, + "loss": 0.0913, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034291915595531464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016848549619317053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08543238118290901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 310, + "weighted_flow_matching_loss": 0.04859849065542221, + "weighted_rkd_loss": 0.042716190591454505 + }, + { + "epoch": 0.5369127516778524, + "flow_matching_loss": 0.04939671419560909, + "grad_norm": 0.4409841299057007, + "learning_rate": 2.998438303643587e-05, + "loss": 0.0928, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03414266463369131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018428620789200066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08671394959092141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 320, + "weighted_flow_matching_loss": 0.04939671419560909, + "weighted_rkd_loss": 0.043356974795460704 + }, + { + "epoch": 0.5536912751677853, + "flow_matching_loss": 0.04955456294119358, + "grad_norm": 0.37391889095306396, + "learning_rate": 2.997952946623527e-05, + "loss": 0.0927, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.033923293277621266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018506798055022956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0863533854484558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 330, + "weighted_flow_matching_loss": 0.04955456294119358, + "weighted_rkd_loss": 0.0431766927242279 + }, + { + "epoch": 0.5704697986577181, + "flow_matching_loss": 0.050061154365539554, + "grad_norm": 0.40724676847457886, + "learning_rate": 2.9974020642888875e-05, + "loss": 0.0919, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03251873888075352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018702817521989344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08374029472470283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 340, + "weighted_flow_matching_loss": 0.050061154365539554, + "weighted_rkd_loss": 0.041870147362351416 + }, + { + "epoch": 0.587248322147651, + "flow_matching_loss": 0.04783603325486183, + "grad_norm": 0.3322409391403198, + "learning_rate": 2.9967856807370458e-05, + "loss": 0.0856, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02946928795427084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016537870932370424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07547644712030888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 350, + "weighted_flow_matching_loss": 0.04783603325486183, + "weighted_rkd_loss": 0.03773822356015444 + }, + { + "epoch": 0.6040268456375839, + "flow_matching_loss": 0.04893294684588909, + "grad_norm": 0.37779635190963745, + "learning_rate": 2.996103822930615e-05, + "loss": 0.0904, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03286735378205776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01721991701051593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08295462429523467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 360, + "weighted_flow_matching_loss": 0.04893294684588909, + "weighted_rkd_loss": 0.04147731214761734 + }, + { + "epoch": 0.6208053691275168, + "flow_matching_loss": 0.04373093023896217, + "grad_norm": 0.3719472587108612, + "learning_rate": 2.9953565206962653e-05, + "loss": 0.0815, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030151859484612943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015161976683884859, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07546569555997848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 370, + "weighted_flow_matching_loss": 0.04373093023896217, + "weighted_rkd_loss": 0.03773284777998924 + }, + { + "epoch": 0.6375838926174496, + "flow_matching_loss": 0.0451244730502367, + "grad_norm": 0.41764843463897705, + "learning_rate": 2.9945438067234172e-05, + "loss": 0.0847, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031219374760985373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016777731571346523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07921648174524307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 380, + "weighted_flow_matching_loss": 0.0451244730502367, + "weighted_rkd_loss": 0.039608240872621533 + }, + { + "epoch": 0.6543624161073825, + "flow_matching_loss": 0.04234543144702911, + "grad_norm": 0.3251534402370453, + "learning_rate": 2.9936657165628124e-05, + "loss": 0.0805, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03002563174813986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01633982127532363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07639108411967754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 390, + "weighted_flow_matching_loss": 0.04234543144702911, + "weighted_rkd_loss": 0.03819554205983877 + }, + { + "epoch": 0.6711409395973155, + "flow_matching_loss": 0.04180084690451622, + "grad_norm": 0.3378938138484955, + "learning_rate": 2.992722288624959e-05, + "loss": 0.0757, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026612223125994205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014572171401232481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06779661700129509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 400, + "weighted_flow_matching_loss": 0.04180084690451622, + "weighted_rkd_loss": 0.033898308500647546 + }, + { + "epoch": 0.6879194630872483, + "flow_matching_loss": 0.04224371202290058, + "grad_norm": 0.3396628797054291, + "learning_rate": 2.9917135641784515e-05, + "loss": 0.0737, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025039970688521863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0127363421022892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06281628347933292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 410, + "weighted_flow_matching_loss": 0.04224371202290058, + "weighted_rkd_loss": 0.03140814173966646 + }, + { + "epoch": 0.7046979865771812, + "flow_matching_loss": 0.03971978519111872, + "grad_norm": 0.30670905113220215, + "learning_rate": 2.9906395873481655e-05, + "loss": 0.0727, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02627517990767956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013494232017546893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06604459211230278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 420, + "weighted_flow_matching_loss": 0.03971978519111872, + "weighted_rkd_loss": 0.03302229605615139 + }, + { + "epoch": 0.7214765100671141, + "flow_matching_loss": 0.04066108614206314, + "grad_norm": 0.39300358295440674, + "learning_rate": 2.989500405113326e-05, + "loss": 0.0738, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026304544880986214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01363995224237442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06624904237687587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 430, + "weighted_flow_matching_loss": 0.04066108614206314, + "weighted_rkd_loss": 0.033124521188437936 + }, + { + "epoch": 0.738255033557047, + "flow_matching_loss": 0.04158542156219482, + "grad_norm": 0.3305646777153015, + "learning_rate": 2.9882960673054547e-05, + "loss": 0.0734, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025151407718658446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01327982721850276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06358264274895191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 440, + "weighted_flow_matching_loss": 0.04158542156219482, + "weighted_rkd_loss": 0.031791321374475955 + }, + { + "epoch": 0.7550335570469798, + "flow_matching_loss": 0.04144757613539696, + "grad_norm": 0.34595179557800293, + "learning_rate": 2.9870266266061886e-05, + "loss": 0.0696, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022307384945452213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0116576106287539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0562723807990551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 450, + "weighted_flow_matching_loss": 0.04144757613539696, + "weighted_rkd_loss": 0.02813619039952755 + }, + { + "epoch": 0.7718120805369127, + "flow_matching_loss": 0.04313542582094669, + "grad_norm": 0.3158428370952606, + "learning_rate": 2.9856921385449766e-05, + "loss": 0.0775, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02726253420114517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014133189246058464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0686582576483488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 460, + "weighted_flow_matching_loss": 0.04313542582094669, + "weighted_rkd_loss": 0.0343291288241744 + }, + { + "epoch": 0.7885906040268457, + "flow_matching_loss": 0.042977604642510416, + "grad_norm": 0.3631656765937805, + "learning_rate": 2.9842926614966502e-05, + "loss": 0.0723, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023044148832559584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012475840467959642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05856413841247558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 470, + "weighted_flow_matching_loss": 0.042977604642510416, + "weighted_rkd_loss": 0.02928206920623779 + }, + { + "epoch": 0.8053691275167785, + "flow_matching_loss": 0.03909936882555485, + "grad_norm": 0.29421156644821167, + "learning_rate": 2.982828256678868e-05, + "loss": 0.069, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02391676288098097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0120180937461555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.059851620346307755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 480, + "weighted_flow_matching_loss": 0.03909936882555485, + "weighted_rkd_loss": 0.029925810173153877 + }, + { + "epoch": 0.8221476510067114, + "flow_matching_loss": 0.03855758477002382, + "grad_norm": 0.2879815399646759, + "learning_rate": 2.9812989881494412e-05, + "loss": 0.0692, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02446096483618021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012417732179164887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061339662224054334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 490, + "weighted_flow_matching_loss": 0.03855758477002382, + "weighted_rkd_loss": 0.030669831112027167 + }, + { + "epoch": 0.8389261744966443, + "flow_matching_loss": 0.04289716742932796, + "grad_norm": 0.37124210596084595, + "learning_rate": 2.9797049228035296e-05, + "loss": 0.0738, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0244005236774683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013081625755876302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06188267394900322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 500, + "weighted_flow_matching_loss": 0.04289716742932796, + "weighted_rkd_loss": 0.03094133697450161 + }, + { + "epoch": 0.8557046979865772, + "flow_matching_loss": 0.03983708620071411, + "grad_norm": 0.3242025375366211, + "learning_rate": 2.978046130370715e-05, + "loss": 0.0676, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022115133143961428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011383708892390132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05561397559940815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 510, + "weighted_flow_matching_loss": 0.03983708620071411, + "weighted_rkd_loss": 0.027806987799704076 + }, + { + "epoch": 0.87248322147651, + "flow_matching_loss": 0.03861053548753261, + "grad_norm": 0.30338215827941895, + "learning_rate": 2.9763226834119535e-05, + "loss": 0.066, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021896894834935666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011009396938607096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054803187027573586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 520, + "weighted_flow_matching_loss": 0.03861053548753261, + "weighted_rkd_loss": 0.027401593513786793 + }, + { + "epoch": 0.889261744966443, + "flow_matching_loss": 0.037504769116640094, + "grad_norm": 0.2798364460468292, + "learning_rate": 2.9745346573163963e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019946480728685857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009748974908143282, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04964193664491177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 530, + "weighted_flow_matching_loss": 0.037504769116640094, + "weighted_rkd_loss": 0.024820968322455884 + }, + { + "epoch": 0.9060402684563759, + "flow_matching_loss": 0.03797667548060417, + "grad_norm": 0.25551819801330566, + "learning_rate": 2.9726821302980987e-05, + "loss": 0.0649, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021121531259268522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01155223953537643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05379530191421509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 540, + "weighted_flow_matching_loss": 0.03797667548060417, + "weighted_rkd_loss": 0.026897650957107545 + }, + { + "epoch": 0.9228187919463087, + "flow_matching_loss": 0.037000850029289725, + "grad_norm": 0.31982243061065674, + "learning_rate": 2.9707651833925935e-05, + "loss": 0.0621, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0204544547945261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009314438654109835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05022334791719914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 550, + "weighted_flow_matching_loss": 0.037000850029289725, + "weighted_rkd_loss": 0.02511167395859957 + }, + { + "epoch": 0.9395973154362416, + "flow_matching_loss": 0.038409977406263354, + "grad_norm": 0.32383039593696594, + "learning_rate": 2.968783900453348e-05, + "loss": 0.0618, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01836938913911581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010035208612680434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046773987263441084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 560, + "weighted_flow_matching_loss": 0.038409977406263354, + "weighted_rkd_loss": 0.023386993631720542 + }, + { + "epoch": 0.9563758389261745, + "flow_matching_loss": 0.03884415850043297, + "grad_norm": 0.3668201267719269, + "learning_rate": 2.9667383681480968e-05, + "loss": 0.0609, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01754778316244483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008952949335798622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04404851570725441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 570, + "weighted_flow_matching_loss": 0.03884415850043297, + "weighted_rkd_loss": 0.022024257853627206 + }, + { + "epoch": 0.9731543624161074, + "flow_matching_loss": 0.03858929816633463, + "grad_norm": 0.31969910860061646, + "learning_rate": 2.9646286759550493e-05, + "loss": 0.0617, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018698493484407663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00886254720389843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04625953398644924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 580, + "weighted_flow_matching_loss": 0.03858929816633463, + "weighted_rkd_loss": 0.02312976699322462 + }, + { + "epoch": 0.9899328859060402, + "flow_matching_loss": 0.037984182313084605, + "grad_norm": 0.37169548869132996, + "learning_rate": 2.9624549161589764e-05, + "loss": 0.0646, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020831290166825056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011616774275898934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0532793540507555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 590, + "weighted_flow_matching_loss": 0.037984182313084605, + "weighted_rkd_loss": 0.02663967702537775 + }, + { + "epoch": 1.0067114093959733, + "flow_matching_loss": 0.032401621155440805, + "grad_norm": 0.2810411751270294, + "learning_rate": 2.9602171838471745e-05, + "loss": 0.0606, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02296459171921015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010374387260526419, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05630357041954994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 600, + "weighted_flow_matching_loss": 0.032401621155440805, + "weighted_rkd_loss": 0.02815178520977497 + }, + { + "epoch": 1.023489932885906, + "flow_matching_loss": 0.03286149390041828, + "grad_norm": 0.2521582245826721, + "learning_rate": 2.957915576905303e-05, + "loss": 0.0546, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017456885334104298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008559525944292545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04347329661250114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 610, + "weighted_flow_matching_loss": 0.03286149390041828, + "weighted_rkd_loss": 0.02173664830625057 + }, + { + "epoch": 1.0402684563758389, + "flow_matching_loss": 0.03205829132348299, + "grad_norm": 0.3185127377510071, + "learning_rate": 2.9555501960131058e-05, + "loss": 0.0558, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01915504578500986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009150991961359978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04746108390390873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 620, + "weighted_flow_matching_loss": 0.03205829132348299, + "weighted_rkd_loss": 0.023730541951954363 + }, + { + "epoch": 1.0570469798657718, + "flow_matching_loss": 0.03440665490925312, + "grad_norm": 0.2903192341327667, + "learning_rate": 2.9531211446400068e-05, + "loss": 0.0576, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018206587247550487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009964327700436116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04637750200927258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 630, + "weighted_flow_matching_loss": 0.03440665490925312, + "weighted_rkd_loss": 0.02318875100463629 + }, + { + "epoch": 1.0738255033557047, + "flow_matching_loss": 0.0327774690464139, + "grad_norm": 0.2796550691127777, + "learning_rate": 2.9506285290405813e-05, + "loss": 0.0552, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01788708958774805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009063597302883863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044837776198983194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 640, + "weighted_flow_matching_loss": 0.0327774690464139, + "weighted_rkd_loss": 0.022418888099491597 + }, + { + "epoch": 1.0906040268456376, + "flow_matching_loss": 0.03253862913697958, + "grad_norm": 0.29687777161598206, + "learning_rate": 2.948072458249911e-05, + "loss": 0.0553, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018407645542174576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008646867936477065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04546215943992138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 650, + "weighted_flow_matching_loss": 0.03253862913697958, + "weighted_rkd_loss": 0.02273107971996069 + }, + { + "epoch": 1.1073825503355705, + "flow_matching_loss": 0.034775618463754654, + "grad_norm": 0.26023975014686584, + "learning_rate": 2.9454530440788125e-05, + "loss": 0.0564, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017261851113289595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008692631544545294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04321633391082287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 660, + "weighted_flow_matching_loss": 0.034775618463754654, + "weighted_rkd_loss": 0.021608166955411436 + }, + { + "epoch": 1.1241610738255035, + "flow_matching_loss": 0.03697704430669546, + "grad_norm": 0.2973390817642212, + "learning_rate": 2.9427704011089474e-05, + "loss": 0.0586, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017315912712365388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008650409011170267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043282233737409116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 670, + "weighted_flow_matching_loss": 0.03697704430669546, + "weighted_rkd_loss": 0.021641116868704558 + }, + { + "epoch": 1.1409395973154361, + "flow_matching_loss": 0.034938321262598035, + "grad_norm": 0.33880263566970825, + "learning_rate": 2.94002464668781e-05, + "loss": 0.0571, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01774648055434227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008731141500175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04422410242259502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 680, + "weighted_flow_matching_loss": 0.034938321262598035, + "weighted_rkd_loss": 0.02211205121129751 + }, + { + "epoch": 1.157718120805369, + "flow_matching_loss": 0.03397301379591226, + "grad_norm": 0.30417579412460327, + "learning_rate": 2.9372159009235936e-05, + "loss": 0.0551, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016883174236863852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0084213403519243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04218768924474716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 690, + "weighted_flow_matching_loss": 0.03397301379591226, + "weighted_rkd_loss": 0.02109384462237358 + }, + { + "epoch": 1.174496644295302, + "flow_matching_loss": 0.033245801366865634, + "grad_norm": 0.2966938614845276, + "learning_rate": 2.9343442866799365e-05, + "loss": 0.0546, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01730244169011712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008166278852149845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04277116190642118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 700, + "weighted_flow_matching_loss": 0.033245801366865634, + "weighted_rkd_loss": 0.02138558095321059 + }, + { + "epoch": 1.191275167785235, + "flow_matching_loss": 0.032241361774504186, + "grad_norm": 0.3242097496986389, + "learning_rate": 2.9314099295705495e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01751971701160073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008714311430230737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04375374540686607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 710, + "weighted_flow_matching_loss": 0.032241361774504186, + "weighted_rkd_loss": 0.021876872703433036 + }, + { + "epoch": 1.2080536912751678, + "flow_matching_loss": 0.030840910412371157, + "grad_norm": 0.3416939675807953, + "learning_rate": 2.928412957953718e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016577553190290927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008127822168171405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041282928362488744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 720, + "weighted_flow_matching_loss": 0.030840910412371157, + "weighted_rkd_loss": 0.020641464181244372 + }, + { + "epoch": 1.2248322147651007, + "flow_matching_loss": 0.030589586496353148, + "grad_norm": 0.3434118926525116, + "learning_rate": 2.925353502926689e-05, + "loss": 0.0487, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014619570132344962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00703063509427011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03626977521926165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 730, + "weighted_flow_matching_loss": 0.030589586496353148, + "weighted_rkd_loss": 0.018134887609630824 + }, + { + "epoch": 1.2416107382550337, + "flow_matching_loss": 0.028109255991876126, + "grad_norm": 0.2770143449306488, + "learning_rate": 2.922231698319937e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015383913926780225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007206043181940913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03797387108206749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 740, + "weighted_flow_matching_loss": 0.028109255991876126, + "weighted_rkd_loss": 0.018986935541033746 + }, + { + "epoch": 1.2583892617449663, + "flow_matching_loss": 0.03261192515492439, + "grad_norm": 0.271056592464447, + "learning_rate": 2.9190476806913088e-05, + "loss": 0.0533, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016976217925548553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007505776965990663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04145821258425712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 750, + "weighted_flow_matching_loss": 0.03261192515492439, + "weighted_rkd_loss": 0.02072910629212856 + }, + { + "epoch": 1.2751677852348993, + "flow_matching_loss": 0.030070590786635876, + "grad_norm": 0.2317705899477005, + "learning_rate": 2.9158015893200504e-05, + "loss": 0.05, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016096030082553624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007664719317108393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03985677920281887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 760, + "weighted_flow_matching_loss": 0.030070590786635876, + "weighted_rkd_loss": 0.019928389601409434 + }, + { + "epoch": 1.2919463087248322, + "flow_matching_loss": 0.033119112625718115, + "grad_norm": 0.2506316006183624, + "learning_rate": 2.9124935662007137e-05, + "loss": 0.0531, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01613023495301604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007737922063097358, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03999839182943106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 770, + "weighted_flow_matching_loss": 0.033119112625718115, + "weighted_rkd_loss": 0.01999919591471553 + }, + { + "epoch": 1.308724832214765, + "flow_matching_loss": 0.031553388386964795, + "grad_norm": 0.2889447808265686, + "learning_rate": 2.9091237560369467e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01787243187427521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009260696731507779, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04500555992126465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 780, + "weighted_flow_matching_loss": 0.031553388386964795, + "weighted_rkd_loss": 0.022502779960632324 + }, + { + "epoch": 1.325503355704698, + "flow_matching_loss": 0.03222518637776375, + "grad_norm": 0.2785235047340393, + "learning_rate": 2.905692306235163e-05, + "loss": 0.0517, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015721357706934214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007582925446331501, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039025640860199926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 790, + "weighted_flow_matching_loss": 0.03222518637776375, + "weighted_rkd_loss": 0.019512820430099963 + }, + { + "epoch": 1.342281879194631, + "flow_matching_loss": 0.030334611982107164, + "grad_norm": 0.31459397077560425, + "learning_rate": 2.9021993668980925e-05, + "loss": 0.0495, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01560999769717455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007159262569621206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03837925773113966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 800, + "weighted_flow_matching_loss": 0.030334611982107164, + "weighted_rkd_loss": 0.01918962886556983 + }, + { + "epoch": 1.3590604026845639, + "flow_matching_loss": 0.03257175628095865, + "grad_norm": 0.2993733286857605, + "learning_rate": 2.8986450908182185e-05, + "loss": 0.0523, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015799124166369437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00794396479614079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039542213454842566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 810, + "weighted_flow_matching_loss": 0.03257175628095865, + "weighted_rkd_loss": 0.019771106727421283 + }, + { + "epoch": 1.3758389261744965, + "flow_matching_loss": 0.032540197111666204, + "grad_norm": 0.31480374932289124, + "learning_rate": 2.8950296334710902e-05, + "loss": 0.0519, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01558954007923603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007458786852657795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03863786738365889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 820, + "weighted_flow_matching_loss": 0.032540197111666204, + "weighted_rkd_loss": 0.019318933691829444 + }, + { + "epoch": 1.3926174496644295, + "flow_matching_loss": 0.03170314058661461, + "grad_norm": 0.2934443950653076, + "learning_rate": 2.8913531530085248e-05, + "loss": 0.0518, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016132532991468905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007935636956244708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040200702473521234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 830, + "weighted_flow_matching_loss": 0.03170314058661461, + "weighted_rkd_loss": 0.020100351236760617 + }, + { + "epoch": 1.4093959731543624, + "flow_matching_loss": 0.03200640454888344, + "grad_norm": 0.26509004831314087, + "learning_rate": 2.887615810251687e-05, + "loss": 0.0497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014211549051105976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006883570412173867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035306668281555174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 840, + "weighted_flow_matching_loss": 0.03200640454888344, + "weighted_rkd_loss": 0.017653334140777587 + }, + { + "epoch": 1.4261744966442953, + "flow_matching_loss": 0.02827311120927334, + "grad_norm": 0.3237679600715637, + "learning_rate": 2.8838177686840577e-05, + "loss": 0.0495, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017118581011891364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00820444175042212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04244160372763872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 850, + "weighted_flow_matching_loss": 0.02827311120927334, + "weighted_rkd_loss": 0.02122080186381936 + }, + { + "epoch": 1.4429530201342282, + "flow_matching_loss": 0.03121668491512537, + "grad_norm": 0.3042275607585907, + "learning_rate": 2.8799591944442776e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016424584202468395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007744280295446515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040593449398875235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 860, + "weighted_flow_matching_loss": 0.03121668491512537, + "weighted_rkd_loss": 0.020296724699437618 + }, + { + "epoch": 1.4597315436241611, + "flow_matching_loss": 0.030370122380554677, + "grad_norm": 0.3472956120967865, + "learning_rate": 2.8760402563188848e-05, + "loss": 0.0531, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017987641133368017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009500452037900686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045475734397768974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 870, + "weighted_flow_matching_loss": 0.030370122380554677, + "weighted_rkd_loss": 0.022737867198884487 + }, + { + "epoch": 1.476510067114094, + "flow_matching_loss": 0.02861998286098242, + "grad_norm": 0.24505336582660675, + "learning_rate": 2.8720611257349285e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0141552341170609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006439156364649534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03474962469190359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 880, + "weighted_flow_matching_loss": 0.02861998286098242, + "weighted_rkd_loss": 0.017374812345951795 + }, + { + "epoch": 1.4932885906040267, + "flow_matching_loss": 0.03184323962777853, + "grad_norm": 0.2608340084552765, + "learning_rate": 2.868021976752471e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01494277287274599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00734082362614572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037226369604468346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 890, + "weighted_flow_matching_loss": 0.03184323962777853, + "weighted_rkd_loss": 0.018613184802234173 + }, + { + "epoch": 1.5100671140939599, + "flow_matching_loss": 0.03042609766125679, + "grad_norm": 0.333564430475235, + "learning_rate": 2.8639229860569738e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0160973247140646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007541899429634213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03973654881119728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 900, + "weighted_flow_matching_loss": 0.03042609766125679, + "weighted_rkd_loss": 0.01986827440559864 + }, + { + "epoch": 1.5268456375838926, + "flow_matching_loss": 0.03012043833732605, + "grad_norm": 0.24364671111106873, + "learning_rate": 2.8597643329515688e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014172045607119798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006650603329762817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034994694776833056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 910, + "weighted_flow_matching_loss": 0.03012043833732605, + "weighted_rkd_loss": 0.017497347388416528 + }, + { + "epoch": 1.5436241610738255, + "flow_matching_loss": 0.03092307262122631, + "grad_norm": 0.27232593297958374, + "learning_rate": 2.855546199349214e-05, + "loss": 0.0486, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014154891762882472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007071852730587125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03538163620978594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 920, + "weighted_flow_matching_loss": 0.03092307262122631, + "weighted_rkd_loss": 0.01769081810489297 + }, + { + "epoch": 1.5604026845637584, + "flow_matching_loss": 0.027929746173322202, + "grad_norm": 0.33788374066352844, + "learning_rate": 2.85126876976474e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012748553790152073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005962743982672691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031459851376712325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 930, + "weighted_flow_matching_loss": 0.027929746173322202, + "weighted_rkd_loss": 0.015729925688356163 + }, + { + "epoch": 1.5771812080536913, + "flow_matching_loss": 0.028270901925861836, + "grad_norm": 0.21897122263908386, + "learning_rate": 2.8469322313067726e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015070199593901634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007116558589041233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0372569577768445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 940, + "weighted_flow_matching_loss": 0.028270901925861836, + "weighted_rkd_loss": 0.01862847888842225 + }, + { + "epoch": 1.5939597315436242, + "flow_matching_loss": 0.02921714950352907, + "grad_norm": 0.3024495542049408, + "learning_rate": 2.8425367736695522e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011751286219805479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005542754894122481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02904532738029957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 950, + "weighted_flow_matching_loss": 0.02921714950352907, + "weighted_rkd_loss": 0.014522663690149784 + }, + { + "epoch": 1.610738255033557, + "flow_matching_loss": 0.02695902418345213, + "grad_norm": 0.2782779335975647, + "learning_rate": 2.838082589124636e-05, + "loss": 0.0458, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015185461472719907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007375222817063332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03774614538997412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 960, + "weighted_flow_matching_loss": 0.02695902418345213, + "weighted_rkd_loss": 0.01887307269498706 + }, + { + "epoch": 1.62751677852349, + "flow_matching_loss": 0.026085390895605087, + "grad_norm": 0.31670045852661133, + "learning_rate": 2.8335698725124843e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014663305971771479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007240531407296658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03656714335083962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 970, + "weighted_flow_matching_loss": 0.026085390895605087, + "weighted_rkd_loss": 0.01828357167541981 + }, + { + "epoch": 1.6442953020134228, + "flow_matching_loss": 0.02555117029696703, + "grad_norm": 0.2121639996767044, + "learning_rate": 2.8289988212339422e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013465657457709312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0062181116081774235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03314942680299282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 980, + "weighted_flow_matching_loss": 0.02555117029696703, + "weighted_rkd_loss": 0.01657471340149641 + }, + { + "epoch": 1.6610738255033557, + "flow_matching_loss": 0.026020037569105627, + "grad_norm": 0.22167570888996124, + "learning_rate": 2.8243696352415997e-05, + "loss": 0.0455, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01538208145648241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008201855747029186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038966018334031106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 990, + "weighted_flow_matching_loss": 0.026020037569105627, + "weighted_rkd_loss": 0.019483009167015553 + }, + { + "epoch": 1.6778523489932886, + "flow_matching_loss": 0.029859650880098343, + "grad_norm": 0.2646207809448242, + "learning_rate": 2.8196825170310495e-05, + "loss": 0.0455, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012622483912855387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005989065254107117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031234032846987246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1000, + "weighted_flow_matching_loss": 0.029859650880098343, + "weighted_rkd_loss": 0.015617016423493623 + }, + { + "epoch": 1.6946308724832215, + "flow_matching_loss": 0.029160509817302227, + "grad_norm": 0.29436981678009033, + "learning_rate": 2.8149376716320256e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014717372134327888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006617354042828083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036052098497748375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1010, + "weighted_flow_matching_loss": 0.029160509817302227, + "weighted_rkd_loss": 0.018026049248874187 + }, + { + "epoch": 1.7114093959731544, + "flow_matching_loss": 0.02514624744653702, + "grad_norm": 0.2813986837863922, + "learning_rate": 2.8101353065994373e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014104087930172681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007160045579075814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03536822162568569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1020, + "weighted_flow_matching_loss": 0.02514624744653702, + "weighted_rkd_loss": 0.017684110812842845 + }, + { + "epoch": 1.7281879194630871, + "flow_matching_loss": 0.025560924783349037, + "grad_norm": 0.24623122811317444, + "learning_rate": 2.8052756320042885e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013708307594060897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006159724202007055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03357633929699659, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1030, + "weighted_flow_matching_loss": 0.025560924783349037, + "weighted_rkd_loss": 0.016788169648498295 + }, + { + "epoch": 1.7449664429530203, + "flow_matching_loss": 0.02597265727818012, + "grad_norm": 0.24410276114940643, + "learning_rate": 2.8003588604244893e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014219438191503286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006340369349345565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0347792461514473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1040, + "weighted_flow_matching_loss": 0.02597265727818012, + "weighted_rkd_loss": 0.01738962307572365 + }, + { + "epoch": 1.761744966442953, + "flow_matching_loss": 0.02521944120526314, + "grad_norm": 0.24740788340568542, + "learning_rate": 2.7953852069355565e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014595606923103332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0072414461523294445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036432659812271596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1050, + "weighted_flow_matching_loss": 0.02521944120526314, + "weighted_rkd_loss": 0.018216329906135798 + }, + { + "epoch": 1.778523489932886, + "flow_matching_loss": 0.027643476985394953, + "grad_norm": 0.3433672785758972, + "learning_rate": 2.790354889101206e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013499390333890915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006195838330313563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03319461885839701, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1060, + "weighted_flow_matching_loss": 0.027643476985394953, + "weighted_rkd_loss": 0.016597309429198504 + }, + { + "epoch": 1.7953020134228188, + "flow_matching_loss": 0.02857300378382206, + "grad_norm": 0.24258039891719818, + "learning_rate": 2.7852681269638366e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013103808369487524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006252429843880236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03246004665270448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1070, + "weighted_flow_matching_loss": 0.02857300378382206, + "weighted_rkd_loss": 0.01623002332635224 + }, + { + "epoch": 1.8120805369127517, + "flow_matching_loss": 0.025308608263731002, + "grad_norm": 0.25732696056365967, + "learning_rate": 2.780125143034902e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014380837325006723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006497780932113529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03525945525616407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1080, + "weighted_flow_matching_loss": 0.025308608263731002, + "weighted_rkd_loss": 0.017629727628082036 + }, + { + "epoch": 1.8288590604026846, + "flow_matching_loss": 0.026384789682924747, + "grad_norm": 0.32120704650878906, + "learning_rate": 2.7749261622851812e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01262503294274211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006175780994817614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031425846926867965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1090, + "weighted_flow_matching_loss": 0.026384789682924747, + "weighted_rkd_loss": 0.015712923463433982 + }, + { + "epoch": 1.8456375838926173, + "flow_matching_loss": 0.027396404929459096, + "grad_norm": 0.21401037275791168, + "learning_rate": 2.7696714121349335e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011610002536326647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005550801800563931, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028770807012915612, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1100, + "weighted_flow_matching_loss": 0.027396404929459096, + "weighted_rkd_loss": 0.014385403506457806 + }, + { + "epoch": 1.8624161073825505, + "flow_matching_loss": 0.0279746038839221, + "grad_norm": 0.20054933428764343, + "learning_rate": 2.764361122443954e-05, + "loss": 0.0424, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011772450804710389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005359048675745726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028903950192034244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1110, + "weighted_flow_matching_loss": 0.0279746038839221, + "weighted_rkd_loss": 0.014451975096017122 + }, + { + "epoch": 1.8791946308724832, + "flow_matching_loss": 0.02448360715061426, + "grad_norm": 0.26301416754722595, + "learning_rate": 2.758995525501517e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011649888660758733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00557361776009202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02887339498847723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1120, + "weighted_flow_matching_loss": 0.02448360715061426, + "weighted_rkd_loss": 0.014436697494238615 + }, + { + "epoch": 1.895973154362416, + "flow_matching_loss": 0.027417019382119177, + "grad_norm": 0.22313368320465088, + "learning_rate": 2.753574856016215e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012046439200639724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005405143462121486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029498021490871905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1130, + "weighted_flow_matching_loss": 0.027417019382119177, + "weighted_rkd_loss": 0.014749010745435952 + }, + { + "epoch": 1.912751677852349, + "flow_matching_loss": 0.02374915238469839, + "grad_norm": 0.2883581817150116, + "learning_rate": 2.7480993511056913e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013785356283187866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063822722062468525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03395298458635807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1140, + "weighted_flow_matching_loss": 0.02374915238469839, + "weighted_rkd_loss": 0.016976492293179036 + }, + { + "epoch": 1.929530201342282, + "flow_matching_loss": 0.024885526299476622, + "grad_norm": 0.2205721139907837, + "learning_rate": 2.7425692502862684e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012991191726177931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006324190460145473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032306574285030365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1150, + "weighted_flow_matching_loss": 0.024885526299476622, + "weighted_rkd_loss": 0.016153287142515182 + }, + { + "epoch": 1.9463087248322148, + "flow_matching_loss": 0.027140559628605843, + "grad_norm": 0.32631516456604004, + "learning_rate": 2.7369847954624728e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01370504880324006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006441503390669822, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03385160099714994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1160, + "weighted_flow_matching_loss": 0.027140559628605843, + "weighted_rkd_loss": 0.01692580049857497 + }, + { + "epoch": 1.9630872483221475, + "flow_matching_loss": 0.027510961145162584, + "grad_norm": 0.19017235934734344, + "learning_rate": 2.731346230916448e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010960731562227011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051203057635575535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02704176902770996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1170, + "weighted_flow_matching_loss": 0.027510961145162584, + "weighted_rkd_loss": 0.01352088451385498 + }, + { + "epoch": 1.9798657718120807, + "flow_matching_loss": 0.023988731019198895, + "grad_norm": 0.2913515865802765, + "learning_rate": 2.7256538032972746e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010911648208275438, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053501016926020386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0271733982488513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1180, + "weighted_flow_matching_loss": 0.023988731019198895, + "weighted_rkd_loss": 0.01358669912442565 + }, + { + "epoch": 1.9966442953020134, + "flow_matching_loss": 0.023296417482197285, + "grad_norm": 0.2592824697494507, + "learning_rate": 2.7199077616101773e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01110992943868041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004422898893244565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026642757840454577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1190, + "weighted_flow_matching_loss": 0.023296417482197285, + "weighted_rkd_loss": 0.013321378920227289 + }, + { + "epoch": 2.0134228187919465, + "flow_matching_loss": 0.024605618230998515, + "grad_norm": 0.3102531135082245, + "learning_rate": 2.714108357205634e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01258283476345241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052482510916888716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03041392080485821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1200, + "weighted_flow_matching_loss": 0.024605618230998515, + "weighted_rkd_loss": 0.015206960402429104 + }, + { + "epoch": 2.030201342281879, + "flow_matching_loss": 0.027245542779564858, + "grad_norm": 0.22417199611663818, + "learning_rate": 2.7082558437683803e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01143276859074831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005770996958017349, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028636534698307514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1210, + "weighted_flow_matching_loss": 0.027245542779564858, + "weighted_rkd_loss": 0.014318267349153757 + }, + { + "epoch": 2.046979865771812, + "flow_matching_loss": 0.023954916559159756, + "grad_norm": 0.20610450208187103, + "learning_rate": 2.702350477306315e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011581839248538017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005343518941663206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028507197834551336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1220, + "weighted_flow_matching_loss": 0.023954916559159756, + "weighted_rkd_loss": 0.014253598917275668 + }, + { + "epoch": 2.063758389261745, + "flow_matching_loss": 0.023449451476335526, + "grad_norm": 0.2560037672519684, + "learning_rate": 2.6963925161392963e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010292424075305461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004632242536172271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02521709054708481, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1230, + "weighted_flow_matching_loss": 0.023449451476335526, + "weighted_rkd_loss": 0.012608545273542405 + }, + { + "epoch": 2.0805369127516777, + "flow_matching_loss": 0.02585367038846016, + "grad_norm": 0.23351788520812988, + "learning_rate": 2.6903822208878476e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013421973027288914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005883032223209739, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032726977951824665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1240, + "weighted_flow_matching_loss": 0.02585367038846016, + "weighted_rkd_loss": 0.016363488975912333 + }, + { + "epoch": 2.097315436241611, + "flow_matching_loss": 0.02222144976258278, + "grad_norm": 0.19517207145690918, + "learning_rate": 2.6843198544617535e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010299140820279717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048344668466597795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025432748906314372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1250, + "weighted_flow_matching_loss": 0.02222144976258278, + "weighted_rkd_loss": 0.012716374453157186 + }, + { + "epoch": 2.1140939597315436, + "flow_matching_loss": 0.026821971125900745, + "grad_norm": 0.3056233823299408, + "learning_rate": 2.6782056820485612e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013456127140671015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006366095319390297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03327834941446781, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1260, + "weighted_flow_matching_loss": 0.026821971125900745, + "weighted_rkd_loss": 0.016639174707233906 + }, + { + "epoch": 2.1308724832214767, + "flow_matching_loss": 0.024398398119956255, + "grad_norm": 0.2585718035697937, + "learning_rate": 2.672039971101979e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011472947150468826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004935935512185097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027881830371916293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1270, + "weighted_flow_matching_loss": 0.024398398119956255, + "weighted_rkd_loss": 0.013940915185958146 + }, + { + "epoch": 2.1476510067114094, + "flow_matching_loss": 0.024629255197942256, + "grad_norm": 0.1899724155664444, + "learning_rate": 2.665822991330178e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010962147265672684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004843188123777509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026767482981085777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1280, + "weighted_flow_matching_loss": 0.024629255197942256, + "weighted_rkd_loss": 0.013383741490542889 + }, + { + "epoch": 2.164429530201342, + "flow_matching_loss": 0.02464695330709219, + "grad_norm": 0.19738611578941345, + "learning_rate": 2.6595550146839934e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011003640946000814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048983822576701645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026905664429068565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1290, + "weighted_flow_matching_loss": 0.02464695330709219, + "weighted_rkd_loss": 0.013452832214534282 + }, + { + "epoch": 2.1812080536912752, + "flow_matching_loss": 0.024324053525924684, + "grad_norm": 0.21592102944850922, + "learning_rate": 2.6532363153450287e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011682957503944635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00482977582141757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0281956909224391, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1300, + "weighted_flow_matching_loss": 0.024324053525924684, + "weighted_rkd_loss": 0.01409784546121955 + }, + { + "epoch": 2.197986577181208, + "flow_matching_loss": 0.02110758163034916, + "grad_norm": 0.18565689027309418, + "learning_rate": 2.6468671697136634e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010184981161728501, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004276543972082436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024646506272256376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1310, + "weighted_flow_matching_loss": 0.02110758163034916, + "weighted_rkd_loss": 0.012323253136128188 + }, + { + "epoch": 2.214765100671141, + "flow_matching_loss": 0.024282259680330754, + "grad_norm": 0.2649759352207184, + "learning_rate": 2.640447856396961e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012323899054899811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005656029563397169, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030303827580064537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1320, + "weighted_flow_matching_loss": 0.024282259680330754, + "weighted_rkd_loss": 0.015151913790032268 + }, + { + "epoch": 2.2315436241610738, + "flow_matching_loss": 0.02368467003107071, + "grad_norm": 0.1718354970216751, + "learning_rate": 2.6339786561964805e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010672940826043486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005090259737335146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02643614150583744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1330, + "weighted_flow_matching_loss": 0.02368467003107071, + "weighted_rkd_loss": 0.01321807075291872 + }, + { + "epoch": 2.248322147651007, + "flow_matching_loss": 0.02225311063230038, + "grad_norm": 0.1958433985710144, + "learning_rate": 2.6274598520959966e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010571935120970011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004792783199809492, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02593665346503258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1340, + "weighted_flow_matching_loss": 0.02225311063230038, + "weighted_rkd_loss": 0.01296832673251629 + }, + { + "epoch": 2.2651006711409396, + "flow_matching_loss": 0.022798995673656463, + "grad_norm": 0.2145613580942154, + "learning_rate": 2.6208917292491185e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010735450917854905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005248784064315259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026719686202704908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1350, + "weighted_flow_matching_loss": 0.022798995673656463, + "weighted_rkd_loss": 0.013359843101352454 + }, + { + "epoch": 2.2818791946308723, + "flow_matching_loss": 0.0219331169500947, + "grad_norm": 0.18633633852005005, + "learning_rate": 2.6142745749668176e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011048201052471996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005030419118702412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02712682094424963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1360, + "weighted_flow_matching_loss": 0.0219331169500947, + "weighted_rkd_loss": 0.013563410472124815 + }, + { + "epoch": 2.2986577181208054, + "flow_matching_loss": 0.021648623794317246, + "grad_norm": 0.33190497756004333, + "learning_rate": 2.607608678704859e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01214893334545195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005797731433995068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030095598101615904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1370, + "weighted_flow_matching_loss": 0.021648623794317246, + "weighted_rkd_loss": 0.015047799050807952 + }, + { + "epoch": 2.315436241610738, + "flow_matching_loss": 0.022526911459863186, + "grad_norm": 0.23687943816184998, + "learning_rate": 2.600894332051139e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009939734917134046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004744947282597422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024624417070299387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1380, + "weighted_flow_matching_loss": 0.022526911459863186, + "weighted_rkd_loss": 0.012312208535149693 + }, + { + "epoch": 2.3322147651006713, + "flow_matching_loss": 0.02625675145536661, + "grad_norm": 0.2014767974615097, + "learning_rate": 2.594131828712932e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011044638883322477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052457981742918495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02733507603406906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1390, + "weighted_flow_matching_loss": 0.02625675145536661, + "weighted_rkd_loss": 0.01366753801703453 + }, + { + "epoch": 2.348993288590604, + "flow_matching_loss": 0.023217788338661192, + "grad_norm": 0.30442988872528076, + "learning_rate": 2.5873214645040414e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01133463829755783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005828363331966102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028497640043497086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1400, + "weighted_flow_matching_loss": 0.023217788338661192, + "weighted_rkd_loss": 0.014248820021748543 + }, + { + "epoch": 2.365771812080537, + "flow_matching_loss": 0.023031603731215, + "grad_norm": 0.25246021151542664, + "learning_rate": 2.5804635373318604e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009690780472010375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004492142889648676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023873704113066198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1410, + "weighted_flow_matching_loss": 0.023031603731215, + "weighted_rkd_loss": 0.011936852056533099 + }, + { + "epoch": 2.38255033557047, + "flow_matching_loss": 0.024585946835577487, + "grad_norm": 0.20293375849723816, + "learning_rate": 2.57355834718434e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011088294768705964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044671180192381145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02664370760321617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1420, + "weighted_flow_matching_loss": 0.024585946835577487, + "weighted_rkd_loss": 0.013321853801608085 + }, + { + "epoch": 2.3993288590604025, + "flow_matching_loss": 0.021987343579530715, + "grad_norm": 0.23088401556015015, + "learning_rate": 2.566606196116867e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010794766712933779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004875816567800939, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026465350203216075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1430, + "weighted_flow_matching_loss": 0.021987343579530715, + "weighted_rkd_loss": 0.013232675101608037 + }, + { + "epoch": 2.4161073825503356, + "flow_matching_loss": 0.024937476590275766, + "grad_norm": 0.21601395308971405, + "learning_rate": 2.559607388239051e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01238170089200139, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005331674730405212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03009507693350315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1440, + "weighted_flow_matching_loss": 0.024937476590275766, + "weighted_rkd_loss": 0.015047538466751576 + }, + { + "epoch": 2.4328859060402683, + "flow_matching_loss": 0.020688330102711917, + "grad_norm": 0.22601929306983948, + "learning_rate": 2.552562229701422e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010842163767665624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046914838254451755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02637581154704094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1450, + "weighted_flow_matching_loss": 0.020688330102711917, + "weighted_rkd_loss": 0.01318790577352047 + }, + { + "epoch": 2.4496644295302015, + "flow_matching_loss": 0.02394238654524088, + "grad_norm": 0.2014225870370865, + "learning_rate": 2.545471028682036e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009968520933762192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004925135779194534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024862177856266498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1460, + "weighted_flow_matching_loss": 0.02394238654524088, + "weighted_rkd_loss": 0.012431088928133249 + }, + { + "epoch": 2.466442953020134, + "flow_matching_loss": 0.02508413475006819, + "grad_norm": 0.28125983476638794, + "learning_rate": 2.538334095373e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011306649027392267, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005316834663972259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027930132672190668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1470, + "weighted_flow_matching_loss": 0.02508413475006819, + "weighted_rkd_loss": 0.013965066336095334 + }, + { + "epoch": 2.4832214765100673, + "flow_matching_loss": 0.023891801387071608, + "grad_norm": 0.29100310802459717, + "learning_rate": 2.531151741966896e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01073241732083261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005184279289096594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026649114117026328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1480, + "weighted_flow_matching_loss": 0.023891801387071608, + "weighted_rkd_loss": 0.013324557058513164 + }, + { + "epoch": 2.5, + "flow_matching_loss": 0.02415292225778103, + "grad_norm": 0.22286556661128998, + "learning_rate": 2.5239242826431302e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010552192945033312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004753443133085966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025857829488813877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1490, + "weighted_flow_matching_loss": 0.02415292225778103, + "weighted_rkd_loss": 0.012928914744406939 + }, + { + "epoch": 2.5167785234899327, + "flow_matching_loss": 0.022898601554334163, + "grad_norm": 0.18709388375282288, + "learning_rate": 2.5166520335541865e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01060750326141715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004809713829308749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02602472063153982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1500, + "weighted_flow_matching_loss": 0.022898601554334163, + "weighted_rkd_loss": 0.01301236031576991 + }, + { + "epoch": 2.533557046979866, + "flow_matching_loss": 0.02009684108197689, + "grad_norm": 0.25758233666419983, + "learning_rate": 2.5093353128117996e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010353196039795875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00480845645070076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025514848809689284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1510, + "weighted_flow_matching_loss": 0.02009684108197689, + "weighted_rkd_loss": 0.012757424404844642 + }, + { + "epoch": 2.5503355704697985, + "flow_matching_loss": 0.02130047921091318, + "grad_norm": 0.16162274777889252, + "learning_rate": 2.501974440473037e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009250646131113172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046441263752058145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023145418614149094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1520, + "weighted_flow_matching_loss": 0.02130047921091318, + "weighted_rkd_loss": 0.011572709307074547 + }, + { + "epoch": 2.5671140939597317, + "flow_matching_loss": 0.0252719109877944, + "grad_norm": 0.21416917443275452, + "learning_rate": 2.4945697385263015e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009564307797700167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004446143656969071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023574758879840375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1530, + "weighted_flow_matching_loss": 0.0252719109877944, + "weighted_rkd_loss": 0.011787379439920188 + }, + { + "epoch": 2.5838926174496644, + "flow_matching_loss": 0.022078343108296393, + "grad_norm": 0.21463052928447723, + "learning_rate": 2.487121530877243e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011745102750137449, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005604003882035613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02909420933574438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1540, + "weighted_flow_matching_loss": 0.022078343108296393, + "weighted_rkd_loss": 0.01454710466787219 + }, + { + "epoch": 2.600671140939597, + "flow_matching_loss": 0.021502092573791744, + "grad_norm": 0.23142777383327484, + "learning_rate": 2.4796301433345943e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0092654038220644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041564338142052295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022687241435050964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1550, + "weighted_flow_matching_loss": 0.021502092573791744, + "weighted_rkd_loss": 0.011343620717525482 + }, + { + "epoch": 2.61744966442953, + "flow_matching_loss": 0.022588416188955306, + "grad_norm": 0.2171335220336914, + "learning_rate": 2.4720959035959148e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010664369678124785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004696646006777883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02602538522332907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1560, + "weighted_flow_matching_loss": 0.022588416188955306, + "weighted_rkd_loss": 0.013012692611664534 + }, + { + "epoch": 2.6342281879194633, + "flow_matching_loss": 0.022199026308953762, + "grad_norm": 0.2863241732120514, + "learning_rate": 2.464519141233257e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011123205535113811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050924249924719335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027338836155831815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1570, + "weighted_flow_matching_loss": 0.022199026308953762, + "weighted_rkd_loss": 0.013669418077915907 + }, + { + "epoch": 2.651006711409396, + "flow_matching_loss": 0.02178021539002657, + "grad_norm": 0.23810994625091553, + "learning_rate": 2.456900187678754e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010764320380985737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005336681194603443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026865322329103946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1580, + "weighted_flow_matching_loss": 0.02178021539002657, + "weighted_rkd_loss": 0.013432661164551973 + }, + { + "epoch": 2.6677852348993287, + "flow_matching_loss": 0.022027206886559726, + "grad_norm": 0.17385552823543549, + "learning_rate": 2.4492393762101147e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008514114422723651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004043441731482744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021071670670062305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1590, + "weighted_flow_matching_loss": 0.022027206886559726, + "weighted_rkd_loss": 0.010535835335031152 + }, + { + "epoch": 2.684563758389262, + "flow_matching_loss": 0.022792240604758263, + "grad_norm": 0.20645810663700104, + "learning_rate": 2.4415370419360508e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00922980452887714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004303267924115062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02276287693530321, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1600, + "weighted_flow_matching_loss": 0.022792240604758263, + "weighted_rkd_loss": 0.011381438467651606 + }, + { + "epoch": 2.7013422818791946, + "flow_matching_loss": 0.021514329593628645, + "grad_norm": 0.2190205603837967, + "learning_rate": 2.4337935217816157e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009895919263362885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00468641291372478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02447825139388442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1610, + "weighted_flow_matching_loss": 0.021514329593628645, + "weighted_rkd_loss": 0.01223912569694221 + }, + { + "epoch": 2.7181208053691277, + "flow_matching_loss": 0.024532474111765623, + "grad_norm": 0.3014610707759857, + "learning_rate": 2.4260091544734665e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010453154426068067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046753128292039035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02558162175118923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1620, + "weighted_flow_matching_loss": 0.024532474111765623, + "weighted_rkd_loss": 0.012790810875594616 + }, + { + "epoch": 2.7348993288590604, + "flow_matching_loss": 0.022621615324169397, + "grad_norm": 0.20125490427017212, + "learning_rate": 2.4181842805250466e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008896962832659483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004234325815923512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022028251271694898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1630, + "weighted_flow_matching_loss": 0.022621615324169397, + "weighted_rkd_loss": 0.011014125635847449 + }, + { + "epoch": 2.751677852348993, + "flow_matching_loss": 0.021766430512070657, + "grad_norm": 0.21964611113071442, + "learning_rate": 2.4103192422216923e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009006364457309245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003936750791035592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021949479822069407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1640, + "weighted_flow_matching_loss": 0.021766430512070657, + "weighted_rkd_loss": 0.010974739911034703 + }, + { + "epoch": 2.7684563758389262, + "flow_matching_loss": 0.022125181183218955, + "grad_norm": 0.18881672620773315, + "learning_rate": 2.402414383605658e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009801744995638729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004149720002897084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023753209877759217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1650, + "weighted_flow_matching_loss": 0.022125181183218955, + "weighted_rkd_loss": 0.011876604938879608 + }, + { + "epoch": 2.785234899328859, + "flow_matching_loss": 0.022959327884018423, + "grad_norm": 0.20707227289676666, + "learning_rate": 2.3944700504610674e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010146900825202466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004097336274571717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024391137808561326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1660, + "weighted_flow_matching_loss": 0.022959327884018423, + "weighted_rkd_loss": 0.012195568904280663 + }, + { + "epoch": 2.802013422818792, + "flow_matching_loss": 0.02108875922858715, + "grad_norm": 0.20151439309120178, + "learning_rate": 2.3864865902987883e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008188423234969378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003603689349256456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019980535749346016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1670, + "weighted_flow_matching_loss": 0.02108875922858715, + "weighted_rkd_loss": 0.009990267874673008 + }, + { + "epoch": 2.8187919463087248, + "flow_matching_loss": 0.02431239653378725, + "grad_norm": 0.17307604849338531, + "learning_rate": 2.378464352341231e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00882768682204187, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004448901535943151, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02210427522659302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1680, + "weighted_flow_matching_loss": 0.02431239653378725, + "weighted_rkd_loss": 0.01105213761329651 + }, + { + "epoch": 2.8355704697986575, + "flow_matching_loss": 0.021007327269762753, + "grad_norm": 0.1367725431919098, + "learning_rate": 2.3704036875070728e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008387606404721738, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035082216607406737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020283434353768826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1690, + "weighted_flow_matching_loss": 0.021007327269762753, + "weighted_rkd_loss": 0.010141717176884413 + }, + { + "epoch": 2.8523489932885906, + "flow_matching_loss": 0.022363264858722687, + "grad_norm": 0.21401777863502502, + "learning_rate": 2.362304948395906e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008863104181364179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004238870576955378, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0219650786370039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1700, + "weighted_flow_matching_loss": 0.022363264858722687, + "weighted_rkd_loss": 0.01098253931850195 + }, + { + "epoch": 2.8691275167785237, + "flow_matching_loss": 0.021167028322815896, + "grad_norm": 0.2148161679506302, + "learning_rate": 2.3541684892728157e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008788073668256402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004058226011693478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021634373534470796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1710, + "weighted_flow_matching_loss": 0.021167028322815896, + "weighted_rkd_loss": 0.010817186767235398 + }, + { + "epoch": 2.8859060402684564, + "flow_matching_loss": 0.02436531288549304, + "grad_norm": 0.21661841869354248, + "learning_rate": 2.345994666052883e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010374712059274316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004931548284366727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025680972170084714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1720, + "weighted_flow_matching_loss": 0.02436531288549304, + "weighted_rkd_loss": 0.012840486085042357 + }, + { + "epoch": 2.902684563758389, + "flow_matching_loss": 0.022530085407197475, + "grad_norm": 0.17632728815078735, + "learning_rate": 2.337783836285614e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008637029910460114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039325655903667215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021206625550985337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1730, + "weighted_flow_matching_loss": 0.022530085407197475, + "weighted_rkd_loss": 0.010603312775492668 + }, + { + "epoch": 2.9194630872483223, + "flow_matching_loss": 0.020638199150562288, + "grad_norm": 0.21533295512199402, + "learning_rate": 2.3295363591393026e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009914667624980212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004476642981171608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024305978044867515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1740, + "weighted_flow_matching_loss": 0.020638199150562288, + "weighted_rkd_loss": 0.012152989022433757 + }, + { + "epoch": 2.936241610738255, + "flow_matching_loss": 0.022945385053753854, + "grad_norm": 0.24226795136928558, + "learning_rate": 2.3212525953853174e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009643034590408207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005037153465673327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024323222693055868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1750, + "weighted_flow_matching_loss": 0.022945385053753854, + "weighted_rkd_loss": 0.012161611346527934 + }, + { + "epoch": 2.953020134228188, + "flow_matching_loss": 0.021384773962199687, + "grad_norm": 0.20606893301010132, + "learning_rate": 2.3129329073823206e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008418774185702204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036228498327545823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020460398215800524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1760, + "weighted_flow_matching_loss": 0.021384773962199687, + "weighted_rkd_loss": 0.010230199107900262 + }, + { + "epoch": 2.969798657718121, + "flow_matching_loss": 0.02141566090285778, + "grad_norm": 0.21554461121559143, + "learning_rate": 2.3045776590604175e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010217978153377771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004908383940346539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025344339944422246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1770, + "weighted_flow_matching_loss": 0.02141566090285778, + "weighted_rkd_loss": 0.012672169972211123 + }, + { + "epoch": 2.9865771812080535, + "flow_matching_loss": 0.02188427047803998, + "grad_norm": 0.23835872113704681, + "learning_rate": 2.296187215905237e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008762435987591743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00426900724414736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021793879009783267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1780, + "weighted_flow_matching_loss": 0.02188427047803998, + "weighted_rkd_loss": 0.010896939504891633 + }, + { + "epoch": 3.0033557046979866, + "flow_matching_loss": 0.022038573306053875, + "grad_norm": 0.1795431524515152, + "learning_rate": 2.2877619449419437e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007520042033866048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003554393211379647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018594477139413357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1790, + "weighted_flow_matching_loss": 0.022038573306053875, + "weighted_rkd_loss": 0.009297238569706679 + }, + { + "epoch": 3.0201342281879193, + "flow_matching_loss": 0.02141010984778404, + "grad_norm": 0.2697470486164093, + "learning_rate": 2.2793022147191843e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009075135085731744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004234813805669546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022385083977133034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1800, + "weighted_flow_matching_loss": 0.02141010984778404, + "weighted_rkd_loss": 0.011192541988566517 + }, + { + "epoch": 3.0369127516778525, + "flow_matching_loss": 0.02096653329208493, + "grad_norm": 0.18201664090156555, + "learning_rate": 2.2708083952929634e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009220260940492154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004256098275072872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022696620132774115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1810, + "weighted_flow_matching_loss": 0.02096653329208493, + "weighted_rkd_loss": 0.011348310066387058 + }, + { + "epoch": 3.053691275167785, + "flow_matching_loss": 0.022351188492029904, + "grad_norm": 0.2090246081352234, + "learning_rate": 2.2622808582104594e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00882717128843069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034880529856309295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02114239577203989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1820, + "weighted_flow_matching_loss": 0.022351188492029904, + "weighted_rkd_loss": 0.010571197886019946 + }, + { + "epoch": 3.0704697986577183, + "flow_matching_loss": 0.020570892561227083, + "grad_norm": 0.18636099994182587, + "learning_rate": 2.2537199764937703e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009009950887411832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004132049717009067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02215195130556822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1830, + "weighted_flow_matching_loss": 0.020570892561227083, + "weighted_rkd_loss": 0.01107597565278411 + }, + { + "epoch": 3.087248322147651, + "flow_matching_loss": 0.02188956905156374, + "grad_norm": 0.18344570696353912, + "learning_rate": 2.245126124623595e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008544707810506225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00419144555926323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021280860900878905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1840, + "weighted_flow_matching_loss": 0.02188956905156374, + "weighted_rkd_loss": 0.010640430450439452 + }, + { + "epoch": 3.1040268456375837, + "flow_matching_loss": 0.022426276095211507, + "grad_norm": 0.2003953754901886, + "learning_rate": 2.2364996785228547e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008748158859089018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003919238899834454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02141555678099394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1850, + "weighted_flow_matching_loss": 0.022426276095211507, + "weighted_rkd_loss": 0.01070777839049697 + }, + { + "epoch": 3.120805369127517, + "flow_matching_loss": 0.02033795341849327, + "grad_norm": 0.22842928767204285, + "learning_rate": 2.227841015540248e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007575100101530552, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034250223077833654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01857522251084447, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1860, + "weighted_flow_matching_loss": 0.02033795341849327, + "weighted_rkd_loss": 0.009287611255422235 + }, + { + "epoch": 3.1375838926174495, + "flow_matching_loss": 0.020844181906431912, + "grad_norm": 0.2263597548007965, + "learning_rate": 2.2191505144337433e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007685949653387069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003535176650620997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018907075840979816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1870, + "weighted_flow_matching_loss": 0.020844181906431912, + "weighted_rkd_loss": 0.009453537920489908 + }, + { + "epoch": 3.1543624161073827, + "flow_matching_loss": 0.019808660540729762, + "grad_norm": 0.18585313856601715, + "learning_rate": 2.210428555354013e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009581155283376575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004591489001177252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023753799498081207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1880, + "weighted_flow_matching_loss": 0.019808660540729762, + "weighted_rkd_loss": 0.011876899749040604 + }, + { + "epoch": 3.1711409395973154, + "flow_matching_loss": 0.020991289615631105, + "grad_norm": 0.22622573375701904, + "learning_rate": 2.201675519827802e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009574074717238545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004285450396127999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023433599807322025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1890, + "weighted_flow_matching_loss": 0.020991289615631105, + "weighted_rkd_loss": 0.011716799903661012 + }, + { + "epoch": 3.1879194630872485, + "flow_matching_loss": 0.02278821486979723, + "grad_norm": 0.22967985272407532, + "learning_rate": 2.1928917907412403e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009182234713807701, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036421044263988734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022006573714315893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1900, + "weighted_flow_matching_loss": 0.02278821486979723, + "weighted_rkd_loss": 0.011003286857157946 + }, + { + "epoch": 3.204697986577181, + "flow_matching_loss": 0.021890855114907026, + "grad_norm": 0.1822008639574051, + "learning_rate": 2.1840777523230937e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008652022993192076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004090635036118328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02139468090608716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1910, + "weighted_flow_matching_loss": 0.021890855114907026, + "weighted_rkd_loss": 0.01069734045304358 + }, + { + "epoch": 3.221476510067114, + "flow_matching_loss": 0.02066585961729288, + "grad_norm": 0.1782473474740982, + "learning_rate": 2.1752337901279552e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00783657394349575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003442372544668615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019115520268678667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1920, + "weighted_flow_matching_loss": 0.02066585961729288, + "weighted_rkd_loss": 0.009557760134339333 + }, + { + "epoch": 3.238255033557047, + "flow_matching_loss": 0.020921053178608418, + "grad_norm": 0.21969790756702423, + "learning_rate": 2.166360291019383e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009407800249755382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038952767383307217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02271087756380439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1930, + "weighted_flow_matching_loss": 0.020921053178608418, + "weighted_rkd_loss": 0.011355438781902195 + }, + { + "epoch": 3.2550335570469797, + "flow_matching_loss": 0.01944717802107334, + "grad_norm": 0.19404731690883636, + "learning_rate": 2.157457643152973e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00837903949432075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037874118657782673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020545491110533477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1940, + "weighted_flow_matching_loss": 0.01944717802107334, + "weighted_rkd_loss": 0.010272745555266739 + }, + { + "epoch": 3.271812080536913, + "flow_matching_loss": 0.0224097715690732, + "grad_norm": 0.17556366324424744, + "learning_rate": 2.148526235959385e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008008029777556658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038434152025729417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01985947471112013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1950, + "weighted_flow_matching_loss": 0.0224097715690732, + "weighted_rkd_loss": 0.009929737355560065 + }, + { + "epoch": 3.2885906040268456, + "flow_matching_loss": 0.019902830384671687, + "grad_norm": 0.2150319367647171, + "learning_rate": 2.1395664601273016e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008881382131949066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041137875756248835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021876551769673826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1960, + "weighted_flow_matching_loss": 0.019902830384671687, + "weighted_rkd_loss": 0.010938275884836913 + }, + { + "epoch": 3.3053691275167787, + "flow_matching_loss": 0.020638061966747046, + "grad_norm": 0.20082297921180725, + "learning_rate": 2.130578707586345e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009990708529949188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004830724210478365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024812141060829164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1970, + "weighted_flow_matching_loss": 0.020638061966747046, + "weighted_rkd_loss": 0.012406070530414582 + }, + { + "epoch": 3.3221476510067114, + "flow_matching_loss": 0.021972791384905578, + "grad_norm": 0.36457595229148865, + "learning_rate": 2.1215633714899264e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008448011986911298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004079858679324389, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020975882653146982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1980, + "weighted_flow_matching_loss": 0.021972791384905578, + "weighted_rkd_loss": 0.010487941326573491 + }, + { + "epoch": 3.3389261744966445, + "flow_matching_loss": 0.020128148049116133, + "grad_norm": 0.2117772102355957, + "learning_rate": 2.112520846198053e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009189584525302052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003930398868396879, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022309568151831626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1990, + "weighted_flow_matching_loss": 0.020128148049116133, + "weighted_rkd_loss": 0.011154784075915813 + }, + { + "epoch": 3.3557046979865772, + "flow_matching_loss": 0.02076246850192547, + "grad_norm": 0.22179049253463745, + "learning_rate": 2.103451527260074e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077252968214452265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034601765451952814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018910769931972028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2000, + "weighted_flow_matching_loss": 0.02076246850192547, + "weighted_rkd_loss": 0.009455384965986014 + }, + { + "epoch": 3.37248322147651, + "flow_matching_loss": 0.019000662211328744, + "grad_norm": 0.22441506385803223, + "learning_rate": 2.094355811397381e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007931182719767093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037250424269586802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01958740781992674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2010, + "weighted_flow_matching_loss": 0.019000662211328744, + "weighted_rkd_loss": 0.00979370390996337 + }, + { + "epoch": 3.389261744966443, + "flow_matching_loss": 0.02324447510764003, + "grad_norm": 0.18134300410747528, + "learning_rate": 2.0852340964860522e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007788776373490691, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033203586237505077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018897911440581082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2020, + "weighted_flow_matching_loss": 0.02324447510764003, + "weighted_rkd_loss": 0.009448955720290541 + }, + { + "epoch": 3.4060402684563758, + "flow_matching_loss": 0.02150107007473707, + "grad_norm": 0.20962178707122803, + "learning_rate": 2.076086781539447e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074673259630799295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032554172910749914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018190069310367107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2030, + "weighted_flow_matching_loss": 0.02150107007473707, + "weighted_rkd_loss": 0.009095034655183553 + }, + { + "epoch": 3.422818791946309, + "flow_matching_loss": 0.018979523424059154, + "grad_norm": 0.2038407325744629, + "learning_rate": 2.066914266690755e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008069745684042572, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033081255503930153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019447616953402757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2040, + "weighted_flow_matching_loss": 0.018979523424059154, + "weighted_rkd_loss": 0.009723808476701378 + }, + { + "epoch": 3.4395973154362416, + "flow_matching_loss": 0.018519976176321507, + "grad_norm": 0.2189130187034607, + "learning_rate": 2.0577169531754906e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008641075901687145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003642701031640172, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02092485288158059, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2050, + "weighted_flow_matching_loss": 0.018519976176321507, + "weighted_rkd_loss": 0.010462426440790295 + }, + { + "epoch": 3.4563758389261743, + "flow_matching_loss": 0.023320261761546134, + "grad_norm": 0.19270910322666168, + "learning_rate": 2.048495243313942e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006468802364543081, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003018935560248792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015956540312618017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2060, + "weighted_flow_matching_loss": 0.023320261761546134, + "weighted_rkd_loss": 0.007978270156309008 + }, + { + "epoch": 3.4731543624161074, + "flow_matching_loss": 0.01975588295608759, + "grad_norm": 0.1822284758090973, + "learning_rate": 2.039249540493572e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008690228406339883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00411738664843142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02149784378707409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2070, + "weighted_flow_matching_loss": 0.01975588295608759, + "weighted_rkd_loss": 0.010748921893537045 + }, + { + "epoch": 3.48993288590604, + "flow_matching_loss": 0.01874214904382825, + "grad_norm": 0.1515883505344391, + "learning_rate": 2.0299802491513755e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008525742217898369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038174554938450456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020868940092623234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2080, + "weighted_flow_matching_loss": 0.01874214904382825, + "weighted_rkd_loss": 0.010434470046311617 + }, + { + "epoch": 3.5067114093959733, + "flow_matching_loss": 0.018373808544129133, + "grad_norm": 0.22102390229701996, + "learning_rate": 2.0206877747561828e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007662647403776646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035474081989377735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018872702866792677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2090, + "weighted_flow_matching_loss": 0.018373808544129133, + "weighted_rkd_loss": 0.009436351433396339 + }, + { + "epoch": 3.523489932885906, + "flow_matching_loss": 0.017856397293508054, + "grad_norm": 0.15369464457035065, + "learning_rate": 2.011372523790927e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008157176710665227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034173124935477973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019731666054576637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2100, + "weighted_flow_matching_loss": 0.017856397293508054, + "weighted_rkd_loss": 0.009865833027288318 + }, + { + "epoch": 3.540268456375839, + "flow_matching_loss": 0.01998805645853281, + "grad_norm": 0.1758951097726822, + "learning_rate": 2.0020349037348643e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006989141134545207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031998127698898314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017178095038980247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2110, + "weighted_flow_matching_loss": 0.01998805645853281, + "weighted_rkd_loss": 0.008589047519490123 + }, + { + "epoch": 3.557046979865772, + "flow_matching_loss": 0.018932559806853533, + "grad_norm": 0.17690835893154144, + "learning_rate": 1.9926753230457443e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007896357215940952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036027796799317003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0193954941816628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2120, + "weighted_flow_matching_loss": 0.018932559806853533, + "weighted_rkd_loss": 0.0096977470908314 + }, + { + "epoch": 3.573825503355705, + "flow_matching_loss": 0.023601194098591804, + "grad_norm": 0.19478943943977356, + "learning_rate": 1.983294191141948e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009023548383265734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004056708514690399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02210380509495735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2130, + "weighted_flow_matching_loss": 0.023601194098591804, + "weighted_rkd_loss": 0.011051902547478676 + }, + { + "epoch": 3.5906040268456376, + "flow_matching_loss": 0.020795693341642617, + "grad_norm": 0.15999910235404968, + "learning_rate": 1.973891918384575e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008007427817210555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003208912722766399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01922376835718751, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2140, + "weighted_flow_matching_loss": 0.020795693341642617, + "weighted_rkd_loss": 0.009611884178593755 + }, + { + "epoch": 3.6073825503355703, + "flow_matching_loss": 0.021222106367349624, + "grad_norm": 0.21778394281864166, + "learning_rate": 1.9644689160594945e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00793103752657771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003410175652243197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01927225077524781, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2150, + "weighted_flow_matching_loss": 0.021222106367349624, + "weighted_rkd_loss": 0.009636125387623906 + }, + { + "epoch": 3.6241610738255035, + "flow_matching_loss": 0.017535629961639643, + "grad_norm": 0.21843908727169037, + "learning_rate": 1.9550255963593553e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007544249249622226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029765219427645207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018065020348876713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2160, + "weighted_flow_matching_loss": 0.017535629961639643, + "weighted_rkd_loss": 0.009032510174438357 + }, + { + "epoch": 3.640939597315436, + "flow_matching_loss": 0.019194155745208264, + "grad_norm": 0.20880937576293945, + "learning_rate": 1.9455623723655524e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00752829727716744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031787700252607467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018235364742577077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2170, + "weighted_flow_matching_loss": 0.019194155745208264, + "weighted_rkd_loss": 0.009117682371288538 + }, + { + "epoch": 3.6577181208053693, + "flow_matching_loss": 0.019468952622264623, + "grad_norm": 0.18948209285736084, + "learning_rate": 1.9360796580301606e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007364970911294222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032614518189802767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017991393711417912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2180, + "weighted_flow_matching_loss": 0.019468952622264623, + "weighted_rkd_loss": 0.008995696855708956 + }, + { + "epoch": 3.674496644295302, + "flow_matching_loss": 0.022077844850718974, + "grad_norm": 0.16150549054145813, + "learning_rate": 1.9265778681578245e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00765900406986475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032386504113674163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01855665845796466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2190, + "weighted_flow_matching_loss": 0.022077844850718974, + "weighted_rkd_loss": 0.00927832922898233 + }, + { + "epoch": 3.6912751677852347, + "flow_matching_loss": 0.019230139348655938, + "grad_norm": 0.25615647435188293, + "learning_rate": 1.9170574183876143e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007380595942959189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032563595566898586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01801755167543888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2200, + "weighted_flow_matching_loss": 0.019230139348655938, + "weighted_rkd_loss": 0.00900877583771944 + }, + { + "epoch": 3.708053691275168, + "flow_matching_loss": 0.017652079928666353, + "grad_norm": 0.1817786544561386, + "learning_rate": 1.9075187251748452e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007373360265046358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032471208134666084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017993841134011746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2210, + "weighted_flow_matching_loss": 0.017652079928666353, + "weighted_rkd_loss": 0.008996920567005873 + }, + { + "epoch": 3.7248322147651005, + "flow_matching_loss": 0.01961641861125827, + "grad_norm": 0.17127038538455963, + "learning_rate": 1.8979622057728596e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00780273643322289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003444134397432208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019049607403576373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2220, + "weighted_flow_matching_loss": 0.01961641861125827, + "weighted_rkd_loss": 0.009524803701788187 + }, + { + "epoch": 3.7416107382550337, + "flow_matching_loss": 0.019033155217766762, + "grad_norm": 0.17839007079601288, + "learning_rate": 1.8883882782147748e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00856158221140504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003942649811506271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02106581414118409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2230, + "weighted_flow_matching_loss": 0.019033155217766762, + "weighted_rkd_loss": 0.010532907070592045 + }, + { + "epoch": 3.7583892617449663, + "flow_matching_loss": 0.019040965475142002, + "grad_norm": 0.1941649317741394, + "learning_rate": 1.8787973612951967e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007480817148461938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031261443509720266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01808777851983905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2240, + "weighted_flow_matching_loss": 0.019040965475142002, + "weighted_rkd_loss": 0.009043889259919525 + }, + { + "epoch": 3.7751677852348995, + "flow_matching_loss": 0.019748223666101693, + "grad_norm": 0.1636851280927658, + "learning_rate": 1.8691898745519028e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008198972046375274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003771033138036728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02016897704452276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2250, + "weighted_flow_matching_loss": 0.019748223666101693, + "weighted_rkd_loss": 0.01008448852226138 + }, + { + "epoch": 3.791946308724832, + "flow_matching_loss": 0.022198669519275425, + "grad_norm": 0.22628392279148102, + "learning_rate": 1.8595662382474863e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007077188324183225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031255503417924047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017279926873743533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2260, + "weighted_flow_matching_loss": 0.022198669519275425, + "weighted_rkd_loss": 0.008639963436871767 + }, + { + "epoch": 3.8087248322147653, + "flow_matching_loss": 0.016530673112720252, + "grad_norm": 0.21448732912540436, + "learning_rate": 1.8499268733509766e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007735572382807731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035640710266306996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019035216141492128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2270, + "weighted_flow_matching_loss": 0.016530673112720252, + "weighted_rkd_loss": 0.009517608070746064 + }, + { + "epoch": 3.825503355704698, + "flow_matching_loss": 0.021276569552719594, + "grad_norm": 0.14651033282279968, + "learning_rate": 1.84027220151942e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00765763008967042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031805066973902285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01849576672539115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2280, + "weighted_flow_matching_loss": 0.021276569552719594, + "weighted_rkd_loss": 0.009247883362695575 + }, + { + "epoch": 3.8422818791946307, + "flow_matching_loss": 0.021208910923451185, + "grad_norm": 0.1972508579492569, + "learning_rate": 1.83060264507944e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007044157804921269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029331859084777536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017021501343697308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2290, + "weighted_flow_matching_loss": 0.021208910923451185, + "weighted_rkd_loss": 0.008510750671848654 + }, + { + "epoch": 3.859060402684564, + "flow_matching_loss": 0.018153464980423452, + "grad_norm": 0.1617676019668579, + "learning_rate": 1.82091862700876e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008048683824017643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003274902165867388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019372269883751868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2300, + "weighted_flow_matching_loss": 0.018153464980423452, + "weighted_rkd_loss": 0.009686134941875934 + }, + { + "epoch": 3.8758389261744965, + "flow_matching_loss": 0.019316170644015074, + "grad_norm": 0.20088458061218262, + "learning_rate": 1.8112205709177023e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00764038790948689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031374069745652377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01841818243265152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2310, + "weighted_flow_matching_loss": 0.019316170644015074, + "weighted_rkd_loss": 0.00920909121632576 + }, + { + "epoch": 3.8926174496644297, + "flow_matching_loss": 0.01878544017672539, + "grad_norm": 0.17331726849079132, + "learning_rate": 1.8015089010306585e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006989720230922103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030413487227633597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0170207891613245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2320, + "weighted_flow_matching_loss": 0.01878544017672539, + "weighted_rkd_loss": 0.00851039458066225 + }, + { + "epoch": 3.9093959731543624, + "flow_matching_loss": 0.018327188584953545, + "grad_norm": 0.18315960466861725, + "learning_rate": 1.7917840421675298e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006432220642454922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026728694792836905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015537310857325792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2330, + "weighted_flow_matching_loss": 0.018327188584953545, + "weighted_rkd_loss": 0.007768655428662896 + }, + { + "epoch": 3.926174496644295, + "flow_matching_loss": 0.017242155969142914, + "grad_norm": 0.17186187207698822, + "learning_rate": 1.7820464197251474e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006151645421050489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027259318390861153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015029222704470158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2340, + "weighted_flow_matching_loss": 0.017242155969142914, + "weighted_rkd_loss": 0.007514611352235079 + }, + { + "epoch": 3.942953020134228, + "flow_matching_loss": 0.02237462541088462, + "grad_norm": 0.24216608703136444, + "learning_rate": 1.7722964596586627e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007923036813735962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003261329559609294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01910740341991186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2350, + "weighted_flow_matching_loss": 0.02237462541088462, + "weighted_rkd_loss": 0.00955370170995593 + }, + { + "epoch": 3.959731543624161, + "flow_matching_loss": 0.018751779105514288, + "grad_norm": 0.23497825860977173, + "learning_rate": 1.7625345884629144e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007246668357402087, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031278782756999136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01762121506035328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2360, + "weighted_flow_matching_loss": 0.018751779105514288, + "weighted_rkd_loss": 0.00881060753017664 + }, + { + "epoch": 3.976510067114094, + "flow_matching_loss": 0.01858077123761177, + "grad_norm": 0.20044608414173126, + "learning_rate": 1.7527612331537722e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006888248072937131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002850493707228452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016626989655196666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2370, + "weighted_flow_matching_loss": 0.01858077123761177, + "weighted_rkd_loss": 0.008313494827598333 + }, + { + "epoch": 3.9932885906040267, + "flow_matching_loss": 0.019617028534412384, + "grad_norm": 0.19277773797512054, + "learning_rate": 1.7429768212494593e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007408017152920365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00341418138705194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0182302157394588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2380, + "weighted_flow_matching_loss": 0.019617028534412384, + "weighted_rkd_loss": 0.0091151078697294 + }, + { + "epoch": 4.010067114093959, + "flow_matching_loss": 0.01653716042637825, + "grad_norm": 0.16038158535957336, + "learning_rate": 1.7331817807518497e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006630103243514896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003022903867531568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01628311024978757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2390, + "weighted_flow_matching_loss": 0.01653716042637825, + "weighted_rkd_loss": 0.008141555124893785 + }, + { + "epoch": 4.026845637583893, + "flow_matching_loss": 0.018175854813307524, + "grad_norm": 0.19850464165210724, + "learning_rate": 1.723376540127745e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00760634196922183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034310010029003023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0186436852440238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2400, + "weighted_flow_matching_loss": 0.018175854813307524, + "weighted_rkd_loss": 0.0093218426220119 + }, + { + "epoch": 4.043624161073826, + "flow_matching_loss": 0.019848077557981014, + "grad_norm": 0.16854286193847656, + "learning_rate": 1.7135615282901357e-05, + "loss": 0.0301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008194463048130273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00404913816601038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020438064355403185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2410, + "weighted_flow_matching_loss": 0.019848077557981014, + "weighted_rkd_loss": 0.010219032177701592 + }, + { + "epoch": 4.060402684563758, + "flow_matching_loss": 0.020925424993038177, + "grad_norm": 0.20957912504673004, + "learning_rate": 1.7037371745794358e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006662683980539441, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029504644102416933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016275832336395978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2420, + "weighted_flow_matching_loss": 0.020925424993038177, + "weighted_rkd_loss": 0.008137916168197989 + }, + { + "epoch": 4.077181208053691, + "flow_matching_loss": 0.019630491454154252, + "grad_norm": 0.1673985868692398, + "learning_rate": 1.6939039087447026e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006573117896914482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002810140186920762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01595637621358037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2430, + "weighted_flow_matching_loss": 0.019630491454154252, + "weighted_rkd_loss": 0.007978188106790185 + }, + { + "epoch": 4.093959731543624, + "flow_matching_loss": 0.021586596686393022, + "grad_norm": 0.201747328042984, + "learning_rate": 1.6840621609248397e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007233408000320196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003126069717109203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01759288590401411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2440, + "weighted_flow_matching_loss": 0.021586596686393022, + "weighted_rkd_loss": 0.008796442952007055 + }, + { + "epoch": 4.110738255033557, + "flow_matching_loss": 0.018460760451853277, + "grad_norm": 0.2036897987127304, + "learning_rate": 1.6742123616297794e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007287621404975653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003143525216728449, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017718768026679755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2450, + "weighted_flow_matching_loss": 0.018460760451853277, + "weighted_rkd_loss": 0.008859384013339877 + }, + { + "epoch": 4.12751677852349, + "flow_matching_loss": 0.017445159237831832, + "grad_norm": 0.16198286414146423, + "learning_rate": 1.664354941721651e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006410072930157185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003017538005951792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0158376838080585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2460, + "weighted_flow_matching_loss": 0.017445159237831832, + "weighted_rkd_loss": 0.00791884190402925 + }, + { + "epoch": 4.144295302013423, + "flow_matching_loss": 0.016915331315249204, + "grad_norm": 0.16699963808059692, + "learning_rate": 1.6544903323959355e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005673299496993423, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020931367063894866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013439735677093268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2470, + "weighted_flow_matching_loss": 0.016915331315249204, + "weighted_rkd_loss": 0.006719867838546634 + }, + { + "epoch": 4.1610738255033555, + "flow_matching_loss": 0.020208620838820933, + "grad_norm": 0.215232715010643, + "learning_rate": 1.6446189651626023e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006692374125123024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026895080460235476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01607425631955266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2480, + "weighted_flow_matching_loss": 0.020208620838820933, + "weighted_rkd_loss": 0.00803712815977633 + }, + { + "epoch": 4.177852348993288, + "flow_matching_loss": 0.020926382672041655, + "grad_norm": 0.22336435317993164, + "learning_rate": 1.6347412718272316e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007076423778198659, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003159638389479369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01731248591095209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2490, + "weighted_flow_matching_loss": 0.020926382672041655, + "weighted_rkd_loss": 0.008656242955476045 + }, + { + "epoch": 4.194630872483222, + "flow_matching_loss": 0.016524071246385573, + "grad_norm": 0.1518680602312088, + "learning_rate": 1.624857684472131e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006714969547465444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003003756864927709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01643369598314166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2500, + "weighted_flow_matching_loss": 0.016524071246385573, + "weighted_rkd_loss": 0.00821684799157083 + }, + { + "epoch": 4.2114093959731544, + "flow_matching_loss": 0.018874516151845454, + "grad_norm": 0.17943860590457916, + "learning_rate": 1.6149686354374283e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006851326674222946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002800596947781742, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0165032503195107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2510, + "weighted_flow_matching_loss": 0.018874516151845454, + "weighted_rkd_loss": 0.00825162515975535 + }, + { + "epoch": 4.228187919463087, + "flow_matching_loss": 0.016614831425249575, + "grad_norm": 0.25215211510658264, + "learning_rate": 1.605074557302165e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006883031688630581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028047516010701657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01657081479206681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2520, + "weighted_flow_matching_loss": 0.016614831425249575, + "weighted_rkd_loss": 0.008285407396033406 + }, + { + "epoch": 4.24496644295302, + "flow_matching_loss": 0.01997323101386428, + "grad_norm": 0.20624391734600067, + "learning_rate": 1.59517588286537e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006685473956167698, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00293945602606982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01631040396168828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2530, + "weighted_flow_matching_loss": 0.01997323101386428, + "weighted_rkd_loss": 0.00815520198084414 + }, + { + "epoch": 4.261744966442953, + "flow_matching_loss": 0.02013480234891176, + "grad_norm": 0.1708594560623169, + "learning_rate": 1.5852730451271324e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005825302423909306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002814886881969869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014465491846203804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2540, + "weighted_flow_matching_loss": 0.02013480234891176, + "weighted_rkd_loss": 0.007232745923101902 + }, + { + "epoch": 4.278523489932886, + "flow_matching_loss": 0.017857098765671254, + "grad_norm": 0.16984955966472626, + "learning_rate": 1.5753664772696546e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006671967078000307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027174580260179936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01606139214709401, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2550, + "weighted_flow_matching_loss": 0.017857098765671254, + "weighted_rkd_loss": 0.008030696073547006 + }, + { + "epoch": 4.295302013422819, + "flow_matching_loss": 0.017334267776459456, + "grad_norm": 0.15617665648460388, + "learning_rate": 1.565456612638307e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006426696386188269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002855884563177824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015709277335554363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2560, + "weighted_flow_matching_loss": 0.017334267776459456, + "weighted_rkd_loss": 0.007854638667777181 + }, + { + "epoch": 4.3120805369127515, + "flow_matching_loss": 0.019305859599262475, + "grad_norm": 0.16668863594532013, + "learning_rate": 1.5555438847226732e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00716956895776093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026791229262016713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017018260527402164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2570, + "weighted_flow_matching_loss": 0.019305859599262475, + "weighted_rkd_loss": 0.008509130263701082 + }, + { + "epoch": 4.328859060402684, + "flow_matching_loss": 0.019607679545879365, + "grad_norm": 0.1785610169172287, + "learning_rate": 1.545628727137584e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009049271885305642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035267358762212098, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021625279542058708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2580, + "weighted_flow_matching_loss": 0.019607679545879365, + "weighted_rkd_loss": 0.010812639771029354 + }, + { + "epoch": 4.345637583892618, + "flow_matching_loss": 0.018800497893244027, + "grad_norm": 0.1957562118768692, + "learning_rate": 1.535711573604153e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00803911597467959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003370839380659163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019449071399867533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2590, + "weighted_flow_matching_loss": 0.018800497893244027, + "weighted_rkd_loss": 0.009724535699933767 + }, + { + "epoch": 4.3624161073825505, + "flow_matching_loss": 0.017709657829254866, + "grad_norm": 0.2177783101797104, + "learning_rate": 1.5257928579308032e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006389908026903868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026443781214766206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015424194186925888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2600, + "weighted_flow_matching_loss": 0.017709657829254866, + "weighted_rkd_loss": 0.007712097093462944 + }, + { + "epoch": 4.379194630872483, + "flow_matching_loss": 0.01953520327806473, + "grad_norm": 0.22895203530788422, + "learning_rate": 1.5158730139942902e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008311857981607319, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003941608686000108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02056532446295023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2610, + "weighted_flow_matching_loss": 0.01953520327806473, + "weighted_rkd_loss": 0.010282662231475115 + }, + { + "epoch": 4.395973154362416, + "flow_matching_loss": 0.020596037432551385, + "grad_norm": 0.13484011590480804, + "learning_rate": 1.5059524757207239e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005220752139575779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022500797058455646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01269158385694027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2620, + "weighted_flow_matching_loss": 0.020596037432551385, + "weighted_rkd_loss": 0.006345791928470135 + }, + { + "epoch": 4.412751677852349, + "flow_matching_loss": 0.018211192917078733, + "grad_norm": 0.17946451902389526, + "learning_rate": 1.4960316770665866e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006453538197092712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030796703533269463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015986746829003095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2630, + "weighted_flow_matching_loss": 0.018211192917078733, + "weighted_rkd_loss": 0.007993373414501547 + }, + { + "epoch": 4.429530201342282, + "flow_matching_loss": 0.0171947973780334, + "grad_norm": 0.1605665385723114, + "learning_rate": 1.4861110519997495e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007121807662770152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029179507866501806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017161566112190486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2640, + "weighted_flow_matching_loss": 0.0171947973780334, + "weighted_rkd_loss": 0.008580783056095243 + }, + { + "epoch": 4.446308724832215, + "flow_matching_loss": 0.017047136928886174, + "grad_norm": 0.21463190019130707, + "learning_rate": 1.4761910344804913e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006163895828649402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027821387513540683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01510993055999279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2650, + "weighted_flow_matching_loss": 0.017047136928886174, + "weighted_rkd_loss": 0.007554965279996395 + }, + { + "epoch": 4.4630872483221475, + "flow_matching_loss": 0.019098941888660192, + "grad_norm": 0.13163383305072784, + "learning_rate": 1.466272058442516e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005966320005245507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023630001582205296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0142956402618438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2660, + "weighted_flow_matching_loss": 0.019098941888660192, + "weighted_rkd_loss": 0.0071478201309219 + }, + { + "epoch": 4.47986577181208, + "flow_matching_loss": 0.01778299594298005, + "grad_norm": 0.2292172610759735, + "learning_rate": 1.4563545577739676e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006542552821338176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030771528254263103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01616225838661194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2670, + "weighted_flow_matching_loss": 0.01778299594298005, + "weighted_rkd_loss": 0.00808112919330597 + }, + { + "epoch": 4.496644295302014, + "flow_matching_loss": 0.015495003666728735, + "grad_norm": 0.17960447072982788, + "learning_rate": 1.4464389662984543e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007274207659065723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002846747334115207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017395162582397462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2680, + "weighted_flow_matching_loss": 0.015495003666728735, + "weighted_rkd_loss": 0.008697581291198731 + }, + { + "epoch": 4.5134228187919465, + "flow_matching_loss": 0.016512521728873252, + "grad_norm": 0.18700137734413147, + "learning_rate": 1.4365257177560684e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006523366179317236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025913154473528265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015638047736138104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2690, + "weighted_flow_matching_loss": 0.016512521728873252, + "weighted_rkd_loss": 0.007819023868069052 + }, + { + "epoch": 4.530201342281879, + "flow_matching_loss": 0.018928809091448784, + "grad_norm": 0.1533912718296051, + "learning_rate": 1.426615245784416e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006431228248402476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026095522684045137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015472008753567934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2700, + "weighted_flow_matching_loss": 0.018928809091448784, + "weighted_rkd_loss": 0.007736004376783967 + }, + { + "epoch": 4.546979865771812, + "flow_matching_loss": 0.01595159135758877, + "grad_norm": 0.13736169040203094, + "learning_rate": 1.4167079838996468e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006889982311986387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030951215769164262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01687508588656783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2710, + "weighted_flow_matching_loss": 0.01595159135758877, + "weighted_rkd_loss": 0.008437542943283916 + }, + { + "epoch": 4.563758389261745, + "flow_matching_loss": 0.017109850142151117, + "grad_norm": 0.1820959895849228, + "learning_rate": 1.4068043654774897e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006259463052265346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027393482509069146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01525827432051301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2720, + "weighted_flow_matching_loss": 0.017109850142151117, + "weighted_rkd_loss": 0.007629137160256505 + }, + { + "epoch": 4.580536912751678, + "flow_matching_loss": 0.018156492337584495, + "grad_norm": 0.182673841714859, + "learning_rate": 1.3969048237342992e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066155354026705025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028239200590178372, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01605499079450965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2730, + "weighted_flow_matching_loss": 0.018156492337584495, + "weighted_rkd_loss": 0.008027495397254825 + }, + { + "epoch": 4.597315436241611, + "flow_matching_loss": 0.01978831607848406, + "grad_norm": 0.20100091397762299, + "learning_rate": 1.3870097917081011e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070770631544291975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002857843821402639, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017011970281600952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2740, + "weighted_flow_matching_loss": 0.01978831607848406, + "weighted_rkd_loss": 0.008505985140800476 + }, + { + "epoch": 4.614093959731544, + "flow_matching_loss": 0.0186252910643816, + "grad_norm": 0.18295343220233917, + "learning_rate": 1.3771197022396508e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006896005268208683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002676388737745583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016468399157747626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2750, + "weighted_flow_matching_loss": 0.0186252910643816, + "weighted_rkd_loss": 0.008234199578873813 + }, + { + "epoch": 4.630872483221476, + "flow_matching_loss": 0.017001602426171304, + "grad_norm": 0.18725721538066864, + "learning_rate": 1.3672349879535015e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075272177811712025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00305442416574806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01810885975137353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2760, + "weighted_flow_matching_loss": 0.017001602426171304, + "weighted_rkd_loss": 0.009054429875686765 + }, + { + "epoch": 4.64765100671141, + "flow_matching_loss": 0.01720100613310933, + "grad_norm": 0.15701332688331604, + "learning_rate": 1.3573560812390781e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006909766187891364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026912813424132763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016510813497006893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2770, + "weighted_flow_matching_loss": 0.01720100613310933, + "weighted_rkd_loss": 0.008255406748503447 + }, + { + "epoch": 4.6644295302013425, + "flow_matching_loss": 0.017884473502635955, + "grad_norm": 0.14792506396770477, + "learning_rate": 1.3474834142317635e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006335597159340978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002531678543891758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015202872920781374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2780, + "weighted_flow_matching_loss": 0.017884473502635955, + "weighted_rkd_loss": 0.007601436460390687 + }, + { + "epoch": 4.681208053691275, + "flow_matching_loss": 0.01976926876232028, + "grad_norm": 0.17041365802288055, + "learning_rate": 1.3376174187939955e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057981258258223535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002652096108067781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01424834793433547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2790, + "weighted_flow_matching_loss": 0.01976926876232028, + "weighted_rkd_loss": 0.007124173967167735 + }, + { + "epoch": 4.697986577181208, + "flow_matching_loss": 0.019026348367333412, + "grad_norm": 0.15966573357582092, + "learning_rate": 1.3277585264963761e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060622354969382284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023348112939856947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014459282439202071, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2800, + "weighted_flow_matching_loss": 0.019026348367333412, + "weighted_rkd_loss": 0.0072296412196010355 + }, + { + "epoch": 4.714765100671141, + "flow_matching_loss": 0.017377084400504828, + "grad_norm": 0.22258582711219788, + "learning_rate": 1.3179071685987928e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0056399278808385135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002916206861846149, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014196062553673982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2810, + "weighted_flow_matching_loss": 0.017377084400504828, + "weighted_rkd_loss": 0.007098031276836991 + }, + { + "epoch": 4.731543624161074, + "flow_matching_loss": 0.017920501250773668, + "grad_norm": 0.2010204941034317, + "learning_rate": 1.308063776031554e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005940242158249021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002402682788670063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014283167012035847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2820, + "weighted_flow_matching_loss": 0.017920501250773668, + "weighted_rkd_loss": 0.007141583506017924 + }, + { + "epoch": 4.748322147651007, + "flow_matching_loss": 0.01623355569317937, + "grad_norm": 0.17003734409809113, + "learning_rate": 1.2982287793765389e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005624843528494239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024171904427930714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013666877429932356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2830, + "weighted_flow_matching_loss": 0.01623355569317937, + "weighted_rkd_loss": 0.006833438714966178 + }, + { + "epoch": 4.76510067114094, + "flow_matching_loss": 0.016382606141269206, + "grad_norm": 0.16437426209449768, + "learning_rate": 1.2884026088483616e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005247332667931914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022283923462964594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012723057717084884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2840, + "weighted_flow_matching_loss": 0.016382606141269206, + "weighted_rkd_loss": 0.006361528858542442 + }, + { + "epoch": 4.781879194630872, + "flow_matching_loss": 0.020086504332721234, + "grad_norm": 0.20742954313755035, + "learning_rate": 1.2785856942755543e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006500152428634465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027116223354823887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015711927181109787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2850, + "weighted_flow_matching_loss": 0.020086504332721234, + "weighted_rkd_loss": 0.007855963590554894 + }, + { + "epoch": 4.798657718120805, + "flow_matching_loss": 0.020468197017908096, + "grad_norm": 0.18006539344787598, + "learning_rate": 1.2687784650817616e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006521564652211964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030275082564912735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016070637572556733, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2860, + "weighted_flow_matching_loss": 0.020468197017908096, + "weighted_rkd_loss": 0.008035318786278367 + }, + { + "epoch": 4.815436241610739, + "flow_matching_loss": 0.020298132486641408, + "grad_norm": 0.21230530738830566, + "learning_rate": 1.2589813502669603e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006211647461168468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002518127963412553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014941422827541828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2870, + "weighted_flow_matching_loss": 0.020298132486641408, + "weighted_rkd_loss": 0.007470711413770914 + }, + { + "epoch": 4.832214765100671, + "flow_matching_loss": 0.018210323061794043, + "grad_norm": 0.17627596855163574, + "learning_rate": 1.2491947783886907e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004903101362287998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021681559388525783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011974358558654785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2880, + "weighted_flow_matching_loss": 0.018210323061794043, + "weighted_rkd_loss": 0.005987179279327392 + }, + { + "epoch": 4.848993288590604, + "flow_matching_loss": 0.016133444383740425, + "grad_norm": 0.13107307255268097, + "learning_rate": 1.2394191775433106e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005635178997181356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002382622950244695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013652980979532003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2890, + "weighted_flow_matching_loss": 0.016133444383740425, + "weighted_rkd_loss": 0.006826490489766001 + }, + { + "epoch": 4.865771812080537, + "flow_matching_loss": 0.016500497329980136, + "grad_norm": 0.1938703954219818, + "learning_rate": 1.2296549753472696e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005910794297233224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002559423819184303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014381012320518494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2900, + "weighted_flow_matching_loss": 0.016500497329980136, + "weighted_rkd_loss": 0.007190506160259247 + }, + { + "epoch": 4.882550335570469, + "flow_matching_loss": 0.015015983954071999, + "grad_norm": 0.18545067310333252, + "learning_rate": 1.2199025989184032e-05, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005890693236142397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025859495741315185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014367335941642523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2910, + "weighted_flow_matching_loss": 0.015015983954071999, + "weighted_rkd_loss": 0.007183667970821261 + }, + { + "epoch": 4.899328859060403, + "flow_matching_loss": 0.015698899049311876, + "grad_norm": 0.20254436135292053, + "learning_rate": 1.21016247485725e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064388867001980545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024498644983395935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01532763782888651, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2920, + "weighted_flow_matching_loss": 0.015698899049311876, + "weighted_rkd_loss": 0.007663818914443255 + }, + { + "epoch": 4.916107382550336, + "flow_matching_loss": 0.016660439129918814, + "grad_norm": 0.18414856493473053, + "learning_rate": 1.2004350292283897e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006194327911362052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027924057794734836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015181061439216137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2930, + "weighted_flow_matching_loss": 0.016660439129918814, + "weighted_rkd_loss": 0.007590530719608068 + }, + { + "epoch": 4.932885906040268, + "flow_matching_loss": 0.01687433701008558, + "grad_norm": 0.16476738452911377, + "learning_rate": 1.1907206875418068e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006046470603905618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024416908156126736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014534632116556168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2940, + "weighted_flow_matching_loss": 0.01687433701008558, + "weighted_rkd_loss": 0.007267316058278084 + }, + { + "epoch": 4.949664429530201, + "flow_matching_loss": 0.016659815981984138, + "grad_norm": 0.15063215792179108, + "learning_rate": 1.1810198747342769e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005068508721888065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021296532708220185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012266670726239682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2950, + "weighted_flow_matching_loss": 0.016659815981984138, + "weighted_rkd_loss": 0.006133335363119841 + }, + { + "epoch": 4.966442953020135, + "flow_matching_loss": 0.014483388233929872, + "grad_norm": 0.18998196721076965, + "learning_rate": 1.1713330151507786e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006280959723517299, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027005726704373957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015262492187321185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2960, + "weighted_flow_matching_loss": 0.014483388233929872, + "weighted_rkd_loss": 0.007631246093660593 + }, + { + "epoch": 4.983221476510067, + "flow_matching_loss": 0.016974644828587772, + "grad_norm": 0.1717025190591812, + "learning_rate": 1.161660532525931e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006291400594636798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028575255651958286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015440326556563378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2970, + "weighted_flow_matching_loss": 0.016974644828587772, + "weighted_rkd_loss": 0.007720163278281689 + }, + { + "epoch": 5.0, + "flow_matching_loss": 0.015258622635155917, + "grad_norm": 0.16946367919445038, + "learning_rate": 1.1520028499654584e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00629035490565002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002330243296455592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014910953119397163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2980, + "weighted_flow_matching_loss": 0.015258622635155917, + "weighted_rkd_loss": 0.007455476559698582 + }, + { + "epoch": 5.016778523489933, + "flow_matching_loss": 0.017898726742714643, + "grad_norm": 0.14455287158489227, + "learning_rate": 1.1423603899276833e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053604997927322985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002361709764227271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01308270925655961, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2990, + "weighted_flow_matching_loss": 0.017898726742714643, + "weighted_rkd_loss": 0.006541354628279805 + }, + { + "epoch": 5.033557046979865, + "flow_matching_loss": 0.01671588756144047, + "grad_norm": 0.18688088655471802, + "learning_rate": 1.1327335742050442e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005779878236353397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002441789919976145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01400154661387205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3000, + "weighted_flow_matching_loss": 0.01671588756144047, + "weighted_rkd_loss": 0.007000773306936025 + }, + { + "epoch": 5.050335570469799, + "flow_matching_loss": 0.015817246027290822, + "grad_norm": 0.1989966332912445, + "learning_rate": 1.1231228239056467e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006527934223413467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002765071636531502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015820940118283035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3010, + "weighted_flow_matching_loss": 0.015817246027290822, + "weighted_rkd_loss": 0.007910470059141517 + }, + { + "epoch": 5.067114093959732, + "flow_matching_loss": 0.01937075946480036, + "grad_norm": 0.19405128061771393, + "learning_rate": 1.113528559434845e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006799411680549383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029540552641265094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01655287845060229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3020, + "weighted_flow_matching_loss": 0.01937075946480036, + "weighted_rkd_loss": 0.008276439225301146 + }, + { + "epoch": 5.083892617449664, + "flow_matching_loss": 0.01867743730545044, + "grad_norm": 0.13398496806621552, + "learning_rate": 1.1039512004768476e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006300268415361643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003146882087457925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015747418720275165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3030, + "weighted_flow_matching_loss": 0.01867743730545044, + "weighted_rkd_loss": 0.007873709360137583 + }, + { + "epoch": 5.100671140939597, + "flow_matching_loss": 0.016951379459351303, + "grad_norm": 0.20637381076812744, + "learning_rate": 1.0943911659763612e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0054731245152652265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021580980042926966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013104346767067909, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3040, + "weighted_flow_matching_loss": 0.016951379459351303, + "weighted_rkd_loss": 0.006552173383533954 + }, + { + "epoch": 5.117449664429531, + "flow_matching_loss": 0.01617958638817072, + "grad_norm": 0.17529357969760895, + "learning_rate": 1.084848874120266e-05, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004983928962610662, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002211737749166787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012179595697671176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3050, + "weighted_flow_matching_loss": 0.01617958638817072, + "weighted_rkd_loss": 0.006089797848835588 + }, + { + "epoch": 5.134228187919463, + "flow_matching_loss": 0.01798154441639781, + "grad_norm": 0.1679459512233734, + "learning_rate": 1.0753247423193196e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051181812305003405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002054891409352422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012291253916919232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3060, + "weighted_flow_matching_loss": 0.01798154441639781, + "weighted_rkd_loss": 0.006145626958459616 + }, + { + "epoch": 5.151006711409396, + "flow_matching_loss": 0.016323540639132262, + "grad_norm": 0.20896075665950775, + "learning_rate": 1.0658191871899012e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005931211682036519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027073875186033546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014569810871034861, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3070, + "weighted_flow_matching_loss": 0.016323540639132262, + "weighted_rkd_loss": 0.007284905435517431 + }, + { + "epoch": 5.167785234899329, + "flow_matching_loss": 0.018675233144313096, + "grad_norm": 0.21605846285820007, + "learning_rate": 1.0563326245357852e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005888878158293665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024224749999120832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014200231153517961, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3080, + "weighted_flow_matching_loss": 0.018675233144313096, + "weighted_rkd_loss": 0.007100115576758981 + }, + { + "epoch": 5.184563758389261, + "flow_matching_loss": 0.015594534948468209, + "grad_norm": 0.21472957730293274, + "learning_rate": 1.0468654693299557e-05, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005332688824273646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002128972625359893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012794350273907185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3090, + "weighted_flow_matching_loss": 0.015594534948468209, + "weighted_rkd_loss": 0.006397175136953592 + }, + { + "epoch": 5.201342281879195, + "flow_matching_loss": 0.01680131210014224, + "grad_norm": 0.15244846045970917, + "learning_rate": 1.0374181356964501e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005860675545409322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002612880081869662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014334231335669755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3100, + "weighted_flow_matching_loss": 0.01680131210014224, + "weighted_rkd_loss": 0.007167115667834878 + }, + { + "epoch": 5.218120805369128, + "flow_matching_loss": 0.01609828118234873, + "grad_norm": 0.17836423218250275, + "learning_rate": 1.0279910368922465e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005648004543036223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002437116380315274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013733125384896993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3110, + "weighted_flow_matching_loss": 0.01609828118234873, + "weighted_rkd_loss": 0.006866562692448497 + }, + { + "epoch": 5.23489932885906, + "flow_matching_loss": 0.016079106274992226, + "grad_norm": 0.1312175989151001, + "learning_rate": 1.018584585289186e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006313463742844761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002551958546973765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015178885776549578, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3120, + "weighted_flow_matching_loss": 0.016079106274992226, + "weighted_rkd_loss": 0.007589442888274789 + }, + { + "epoch": 5.251677852348993, + "flow_matching_loss": 0.018048496171832084, + "grad_norm": 0.14827987551689148, + "learning_rate": 1.0091991923559348e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006192484707571566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002579770388547331, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014964739698916674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3130, + "weighted_flow_matching_loss": 0.018048496171832084, + "weighted_rkd_loss": 0.007482369849458337 + }, + { + "epoch": 5.268456375838926, + "flow_matching_loss": 0.016098976973444222, + "grad_norm": 0.21780844032764435, + "learning_rate": 9.99835268639984e-06, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005791219044476747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024931206949986517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014075558725744486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3140, + "weighted_flow_matching_loss": 0.016098976973444222, + "weighted_rkd_loss": 0.007037779362872243 + }, + { + "epoch": 5.285234899328859, + "flow_matching_loss": 0.0159193379804492, + "grad_norm": 0.19609591364860535, + "learning_rate": 9.904932237496913e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004937235242687165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00216703173937276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012041502259671688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3150, + "weighted_flow_matching_loss": 0.0159193379804492, + "weighted_rkd_loss": 0.006020751129835844 + }, + { + "epoch": 5.302013422818792, + "flow_matching_loss": 0.014781306311488152, + "grad_norm": 0.2759782373905182, + "learning_rate": 9.811734663363659e-06, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005513566290028393, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00228801624616608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013315148837864398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3160, + "weighted_flow_matching_loss": 0.014781306311488152, + "weighted_rkd_loss": 0.006657574418932199 + }, + { + "epoch": 5.318791946308725, + "flow_matching_loss": 0.01723882174119353, + "grad_norm": 0.16407190263271332, + "learning_rate": 9.718764040763875e-06, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059700444340705875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022358012502081692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014175890199840068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3170, + "weighted_flow_matching_loss": 0.01723882174119353, + "weighted_rkd_loss": 0.007087945099920034 + }, + { + "epoch": 5.3355704697986575, + "flow_matching_loss": 0.016165380273014306, + "grad_norm": 0.15735647082328796, + "learning_rate": 9.626024436533777e-06, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005720328283496201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002354632946662605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013795289304107427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3180, + "weighted_flow_matching_loss": 0.016165380273014306, + "weighted_rkd_loss": 0.006897644652053714 + }, + { + "epoch": 5.35234899328859, + "flow_matching_loss": 0.01835033930838108, + "grad_norm": 0.18436121940612793, + "learning_rate": 9.533519907404095e-06, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005160203762352466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002100703300675377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012421110738068819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3190, + "weighted_flow_matching_loss": 0.01835033930838108, + "weighted_rkd_loss": 0.006210555369034409 + }, + { + "epoch": 5.369127516778524, + "flow_matching_loss": 0.0171617632266134, + "grad_norm": 0.17887401580810547, + "learning_rate": 9.4412544998226e-06, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004801495070569217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022160978172905744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011819087900221348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3200, + "weighted_flow_matching_loss": 0.0171617632266134, + "weighted_rkd_loss": 0.005909543950110674 + }, + { + "epoch": 5.385906040268456, + "flow_matching_loss": 0.01628105537965894, + "grad_norm": 0.21056130528450012, + "learning_rate": 9.349232249777124e-06, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059770388295874, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024258324061520397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01437990991398692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3210, + "weighted_flow_matching_loss": 0.01628105537965894, + "weighted_rkd_loss": 0.00718995495699346 + }, + { + "epoch": 5.402684563758389, + "flow_matching_loss": 0.019007640797644852, + "grad_norm": 0.2918168008327484, + "learning_rate": 9.257457182618985e-06, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005413648206740618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002455686731263995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013282982911914588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3220, + "weighted_flow_matching_loss": 0.019007640797644852, + "weighted_rkd_loss": 0.006641491455957294 + }, + { + "epoch": 5.419463087248322, + "flow_matching_loss": 0.015169256553053855, + "grad_norm": 0.15156346559524536, + "learning_rate": 9.165933312886944e-06, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004989207885228097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002060519380029291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01203893506899476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3230, + "weighted_flow_matching_loss": 0.015169256553053855, + "weighted_rkd_loss": 0.00601946753449738 + }, + { + "epoch": 5.436241610738255, + "flow_matching_loss": 0.016222430299967526, + "grad_norm": 0.153106227517128, + "learning_rate": 9.07466464413155e-06, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059154522605240345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021968859364278615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014027790259569883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3240, + "weighted_flow_matching_loss": 0.016222430299967526, + "weighted_rkd_loss": 0.0070138951297849415 + }, + { + "epoch": 5.453020134228188, + "flow_matching_loss": 0.016991760395467282, + "grad_norm": 0.23132218420505524, + "learning_rate": 8.983655168740048e-06, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006680940859951079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002693852235097438, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016055733896791935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3250, + "weighted_flow_matching_loss": 0.016991760395467282, + "weighted_rkd_loss": 0.008027866948395967 + }, + { + "epoch": 5.469798657718121, + "flow_matching_loss": 0.016313020419329406, + "grad_norm": 0.14859110116958618, + "learning_rate": 8.892908867761718e-06, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006743181962519884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002991057641338557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016477421671152116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3260, + "weighted_flow_matching_loss": 0.016313020419329406, + "weighted_rkd_loss": 0.008238710835576058 + }, + { + "epoch": 5.4865771812080535, + "flow_matching_loss": 0.018052593991160392, + "grad_norm": 0.16553567349910736, + "learning_rate": 8.802429710733745e-06, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004876571684144437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002291589602828026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012044732971116901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3270, + "weighted_flow_matching_loss": 0.018052593991160392, + "weighted_rkd_loss": 0.006022366485558451 + }, + { + "epoch": 5.503355704697986, + "flow_matching_loss": 0.01667319508269429, + "grad_norm": 0.18746444582939148, + "learning_rate": 8.71222165550757e-06, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004696449846960604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018371382844634353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01123003801330924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3280, + "weighted_flow_matching_loss": 0.01667319508269429, + "weighted_rkd_loss": 0.00561501900665462 + }, + { + "epoch": 5.52013422818792, + "flow_matching_loss": 0.015698192361742257, + "grad_norm": 0.19901099801063538, + "learning_rate": 8.622288648075748e-06, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0054225688334554436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021200297283940016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012965167313814164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3290, + "weighted_flow_matching_loss": 0.015698192361742257, + "weighted_rkd_loss": 0.006482583656907082 + }, + { + "epoch": 5.5369127516778525, + "flow_matching_loss": 0.016813407838344573, + "grad_norm": 0.16385048627853394, + "learning_rate": 8.532634622399391e-06, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004744720063172281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002088817674666643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011578257894143462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3300, + "weighted_flow_matching_loss": 0.016813407838344573, + "weighted_rkd_loss": 0.005789128947071731 + }, + { + "epoch": 5.553691275167785, + "flow_matching_loss": 0.015239028632640839, + "grad_norm": 0.2184711992740631, + "learning_rate": 8.44326350023601e-06, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005157306534238159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025373138021677734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012851926730945706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3310, + "weighted_flow_matching_loss": 0.015239028632640839, + "weighted_rkd_loss": 0.006425963365472853 + }, + { + "epoch": 5.570469798657718, + "flow_matching_loss": 0.015581510215997695, + "grad_norm": 0.15516401827335358, + "learning_rate": 8.354179190968002e-06, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004385465546511114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015736499801278115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01034458102658391, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3320, + "weighted_flow_matching_loss": 0.015581510215997695, + "weighted_rkd_loss": 0.005172290513291955 + }, + { + "epoch": 5.587248322147651, + "flow_matching_loss": 0.018718068767338992, + "grad_norm": 0.21937350928783417, + "learning_rate": 8.265385591431661e-06, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005961455311626196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002444499370176345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014367409981787205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3330, + "weighted_flow_matching_loss": 0.018718068767338992, + "weighted_rkd_loss": 0.007183704990893602 + }, + { + "epoch": 5.604026845637584, + "flow_matching_loss": 0.014946687500923872, + "grad_norm": 0.14376555383205414, + "learning_rate": 8.176886585746671e-06, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004589359811507166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019512920989654958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011130011733621359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3340, + "weighted_flow_matching_loss": 0.014946687500923872, + "weighted_rkd_loss": 0.005565005866810679 + }, + { + "epoch": 5.620805369127517, + "flow_matching_loss": 0.013927073497325183, + "grad_norm": 0.1809721142053604, + "learning_rate": 8.088686045146247e-06, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005342445382848382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022515961551107465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012936487048864364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3350, + "weighted_flow_matching_loss": 0.013927073497325183, + "weighted_rkd_loss": 0.006468243524432182 + }, + { + "epoch": 5.6375838926174495, + "flow_matching_loss": 0.015200373250991106, + "grad_norm": 0.19390113651752472, + "learning_rate": 8.00078782780775e-06, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005480418261140585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002286845410708338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01324768178164959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3360, + "weighted_flow_matching_loss": 0.015200373250991106, + "weighted_rkd_loss": 0.006623840890824795 + }, + { + "epoch": 5.654362416107382, + "flow_matching_loss": 0.017947501968592405, + "grad_norm": 0.1948995292186737, + "learning_rate": 7.913195778683979e-06, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00509554436430335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002246081584598869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012437170092016458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3370, + "weighted_flow_matching_loss": 0.017947501968592405, + "weighted_rkd_loss": 0.006218585046008229 + }, + { + "epoch": 5.671140939597316, + "flow_matching_loss": 0.01727897087112069, + "grad_norm": 0.19190214574337006, + "learning_rate": 7.825913729334908e-06, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005547965504229068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00244416092755273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013540091924369335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3380, + "weighted_flow_matching_loss": 0.01727897087112069, + "weighted_rkd_loss": 0.006770045962184668 + }, + { + "epoch": 5.6879194630872485, + "flow_matching_loss": 0.018374356906861067, + "grad_norm": 0.1317068636417389, + "learning_rate": 7.73894549776013e-06, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0055965505540370945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022137959022074938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013406896963715553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3390, + "weighted_flow_matching_loss": 0.018374356906861067, + "weighted_rkd_loss": 0.006703448481857777 + }, + { + "epoch": 5.704697986577181, + "flow_matching_loss": 0.015805857069790364, + "grad_norm": 0.2368166297674179, + "learning_rate": 7.65229488823184e-06, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053893422009423375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023492945125326513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013127978704869748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3400, + "weighted_flow_matching_loss": 0.015805857069790364, + "weighted_rkd_loss": 0.006563989352434874 + }, + { + "epoch": 5.721476510067114, + "flow_matching_loss": 0.01789247877895832, + "grad_norm": 0.16481158137321472, + "learning_rate": 7.565965691128396e-06, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004735127929598093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020541234756819905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011524379393085838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3410, + "weighted_flow_matching_loss": 0.01789247877895832, + "weighted_rkd_loss": 0.005762189696542919 + }, + { + "epoch": 5.7382550335570475, + "flow_matching_loss": 0.01610357007011771, + "grad_norm": 0.19617056846618652, + "learning_rate": 7.47996168276856e-06, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004507807712070644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017629609443247317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01077857632189989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3420, + "weighted_flow_matching_loss": 0.01610357007011771, + "weighted_rkd_loss": 0.005389288160949945 + }, + { + "epoch": 5.75503355704698, + "flow_matching_loss": 0.016361732501536608, + "grad_norm": 0.16310381889343262, + "learning_rate": 7.394286625246258e-06, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005316798621788621, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018169037997722627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012450501043349504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3430, + "weighted_flow_matching_loss": 0.016361732501536608, + "weighted_rkd_loss": 0.006225250521674752 + }, + { + "epoch": 5.771812080536913, + "flow_matching_loss": 0.01591941826045513, + "grad_norm": 0.14360399544239044, + "learning_rate": 7.308944266266063e-06, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0050241583026945594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00202472114469856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012073037773370742, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3440, + "weighted_flow_matching_loss": 0.01591941826045513, + "weighted_rkd_loss": 0.006036518886685371 + }, + { + "epoch": 5.7885906040268456, + "flow_matching_loss": 0.0171136362478137, + "grad_norm": 0.25523585081100464, + "learning_rate": 7.223938338979233e-06, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0050462506711483005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001889492047484964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011981993541121483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3450, + "weighted_flow_matching_loss": 0.0171136362478137, + "weighted_rkd_loss": 0.005990996770560741 + }, + { + "epoch": 5.805369127516778, + "flow_matching_loss": 0.017496591899544, + "grad_norm": 0.1567636877298355, + "learning_rate": 7.139272561820397e-06, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00436295464169234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016373030026443302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010363212414085865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3460, + "weighted_flow_matching_loss": 0.017496591899544, + "weighted_rkd_loss": 0.005181606207042932 + }, + { + "epoch": 5.822147651006711, + "flow_matching_loss": 0.014734391588717699, + "grad_norm": 0.15370847284793854, + "learning_rate": 7.054950638344936e-06, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0047149185324087736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020497960387729107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01147963311523199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3470, + "weighted_flow_matching_loss": 0.014734391588717699, + "weighted_rkd_loss": 0.005739816557615995 + }, + { + "epoch": 5.8389261744966445, + "flow_matching_loss": 0.016384201217442752, + "grad_norm": 0.18166224658489227, + "learning_rate": 6.970976257066941e-06, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005584712466225028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002454860834404826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013624285720288754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3480, + "weighted_flow_matching_loss": 0.016384201217442752, + "weighted_rkd_loss": 0.006812142860144377 + }, + { + "epoch": 5.855704697986577, + "flow_matching_loss": 0.015255410224199295, + "grad_norm": 0.22488172352313995, + "learning_rate": 6.887353091297896e-06, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004990561841987073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019109268672764302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011892050690948962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3490, + "weighted_flow_matching_loss": 0.015255410224199295, + "weighted_rkd_loss": 0.005946025345474481 + }, + { + "epoch": 5.87248322147651, + "flow_matching_loss": 0.016748811397701503, + "grad_norm": 0.17458610236644745, + "learning_rate": 6.804084798985965e-06, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005321139958687127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001955470233224332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01259775012731552, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3500, + "weighted_flow_matching_loss": 0.016748811397701503, + "weighted_rkd_loss": 0.00629887506365776 + }, + { + "epoch": 5.889261744966443, + "flow_matching_loss": 0.017467737291008235, + "grad_norm": 0.2232614904642105, + "learning_rate": 6.721175022556009e-06, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005391786922700703, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002220256079453975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013003829959779977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3510, + "weighted_flow_matching_loss": 0.017467737291008235, + "weighted_rkd_loss": 0.006501914979889989 + }, + { + "epoch": 5.906040268456376, + "flow_matching_loss": 0.015664891712367533, + "grad_norm": 0.1693728268146515, + "learning_rate": 6.638627388750245e-06, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005543821305036545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001970905240159482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01305854767560959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3520, + "weighted_flow_matching_loss": 0.015664891712367533, + "weighted_rkd_loss": 0.006529273837804795 + }, + { + "epoch": 5.922818791946309, + "flow_matching_loss": 0.018745915219187736, + "grad_norm": 0.1922639161348343, + "learning_rate": 6.556445508469583e-06, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004745194385759532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019202875206246971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011410676408559084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3530, + "weighted_flow_matching_loss": 0.018745915219187736, + "weighted_rkd_loss": 0.005705338204279542 + }, + { + "epoch": 5.939597315436242, + "flow_matching_loss": 0.017241082154214382, + "grad_norm": 0.19612009823322296, + "learning_rate": 6.474632976615708e-06, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005357312108390033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002517049363814294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013231673557311297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3540, + "weighted_flow_matching_loss": 0.017241082154214382, + "weighted_rkd_loss": 0.006615836778655649 + }, + { + "epoch": 5.956375838926174, + "flow_matching_loss": 0.016150518599897624, + "grad_norm": 0.17087864875793457, + "learning_rate": 6.393193371933791e-06, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005487588234245777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025250311009585857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013500207662582397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3550, + "weighted_flow_matching_loss": 0.016150518599897624, + "weighted_rkd_loss": 0.0067501038312911986 + }, + { + "epoch": 5.973154362416107, + "flow_matching_loss": 0.015010569430887698, + "grad_norm": 0.14347855746746063, + "learning_rate": 6.312130256855977e-06, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004399256804026664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019206363591365515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010719149839133024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3560, + "weighted_flow_matching_loss": 0.015010569430887698, + "weighted_rkd_loss": 0.005359574919566512 + }, + { + "epoch": 5.989932885906041, + "flow_matching_loss": 0.015253768581897021, + "grad_norm": 0.1432712972164154, + "learning_rate": 6.23144717734552e-06, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005114776152186096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019915621145628394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012221114430576562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3570, + "weighted_flow_matching_loss": 0.015253768581897021, + "weighted_rkd_loss": 0.006110557215288281 + }, + { + "epoch": 6.006711409395973, + "flow_matching_loss": 0.015888039860874413, + "grad_norm": 0.21380585432052612, + "learning_rate": 6.151147662741695e-06, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004904805589467287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020907162339426576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011900327261537313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3580, + "weighted_flow_matching_loss": 0.015888039860874413, + "weighted_rkd_loss": 0.0059501636307686566 + }, + { + "epoch": 6.023489932885906, + "flow_matching_loss": 0.01624783258885145, + "grad_norm": 0.17643627524375916, + "learning_rate": 6.071235225605415e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004626808851025998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019845670787617562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01123818475753069, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3590, + "weighted_flow_matching_loss": 0.01624783258885145, + "weighted_rkd_loss": 0.005619092378765345 + }, + { + "epoch": 6.040268456375839, + "flow_matching_loss": 0.016807893011718988, + "grad_norm": 0.13113495707511902, + "learning_rate": 5.991713361565546e-06, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00433010421693325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018857970600947738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010546005656942725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3600, + "weighted_flow_matching_loss": 0.016807893011718988, + "weighted_rkd_loss": 0.005273002828471362 + }, + { + "epoch": 6.057046979865772, + "flow_matching_loss": 0.014842659141868353, + "grad_norm": 0.2173433154821396, + "learning_rate": 5.912585549166051e-06, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005738632497377693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024297324125654995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013906997302547098, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3610, + "weighted_flow_matching_loss": 0.014842659141868353, + "weighted_rkd_loss": 0.006953498651273549 + }, + { + "epoch": 6.073825503355705, + "flow_matching_loss": 0.015946427546441554, + "grad_norm": 0.17731747031211853, + "learning_rate": 5.833855249713777e-06, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042516337241977455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017034933145623655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010206760745495558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3620, + "weighted_flow_matching_loss": 0.015946427546441554, + "weighted_rkd_loss": 0.005103380372747779 + }, + { + "epoch": 6.090604026845638, + "flow_matching_loss": 0.017258873023092747, + "grad_norm": 0.18392245471477509, + "learning_rate": 5.7555259071270645e-06, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005527050793170929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023105093743652104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013364610914140939, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3630, + "weighted_flow_matching_loss": 0.017258873023092747, + "weighted_rkd_loss": 0.0066823054570704695 + }, + { + "epoch": 6.10738255033557, + "flow_matching_loss": 0.014458374958485365, + "grad_norm": 0.15555499494075775, + "learning_rate": 5.677600947785117e-06, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004228003416210413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018116092891432346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010267616109922528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3640, + "weighted_flow_matching_loss": 0.014458374958485365, + "weighted_rkd_loss": 0.005133808054961264 + }, + { + "epoch": 6.124161073825503, + "flow_matching_loss": 0.014426036737859249, + "grad_norm": 0.12407287210226059, + "learning_rate": 5.600083780378099e-06, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004010814358480275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015088674030266703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009530496131628752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3650, + "weighted_flow_matching_loss": 0.014426036737859249, + "weighted_rkd_loss": 0.004765248065814376 + }, + { + "epoch": 6.140939597315437, + "flow_matching_loss": 0.015246859192848206, + "grad_norm": 0.19507645070552826, + "learning_rate": 5.522977795758038e-06, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005159773584455252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018581768497824668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012177724111825227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3660, + "weighted_flow_matching_loss": 0.015246859192848206, + "weighted_rkd_loss": 0.0060888620559126135 + }, + { + "epoch": 6.157718120805369, + "flow_matching_loss": 0.016874602250754833, + "grad_norm": 0.1572439819574356, + "learning_rate": 5.4462863667904805e-06, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005385365919210017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023309690644964577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013101700972765685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3670, + "weighted_flow_matching_loss": 0.016874602250754833, + "weighted_rkd_loss": 0.006550850486382842 + }, + { + "epoch": 6.174496644295302, + "flow_matching_loss": 0.017885672207921744, + "grad_norm": 0.17745526134967804, + "learning_rate": 5.370012848206983e-06, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0054987040115520355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002123111695982516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01312051983550191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3680, + "weighted_flow_matching_loss": 0.017885672207921744, + "weighted_rkd_loss": 0.006560259917750955 + }, + { + "epoch": 6.191275167785235, + "flow_matching_loss": 0.014684625109657646, + "grad_norm": 0.2061356008052826, + "learning_rate": 5.29416057645834e-06, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004387064115144312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018910664482973516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010665194783359765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3690, + "weighted_flow_matching_loss": 0.014684625109657646, + "weighted_rkd_loss": 0.005332597391679883 + }, + { + "epoch": 6.208053691275167, + "flow_matching_loss": 0.016030605789273976, + "grad_norm": 0.14315824210643768, + "learning_rate": 5.218732869568635e-06, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051259903470054265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002160561189521104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012412541825324298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3700, + "weighted_flow_matching_loss": 0.016030605789273976, + "weighted_rkd_loss": 0.006206270912662149 + }, + { + "epoch": 6.224832214765101, + "flow_matching_loss": 0.014568359684199094, + "grad_norm": 0.14933782815933228, + "learning_rate": 5.14373302699013e-06, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043758470332249996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017608381458558142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010512532247230411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3710, + "weighted_flow_matching_loss": 0.014568359684199094, + "weighted_rkd_loss": 0.005256266123615206 + }, + { + "epoch": 6.241610738255034, + "flow_matching_loss": 0.015898380894213915, + "grad_norm": 0.15506769716739655, + "learning_rate": 5.069164329458894e-06, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004872710234485567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019389147055335343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011684335116297007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3720, + "weighted_flow_matching_loss": 0.015898380894213915, + "weighted_rkd_loss": 0.005842167558148504 + }, + { + "epoch": 6.258389261744966, + "flow_matching_loss": 0.01669842703267932, + "grad_norm": 0.2566884160041809, + "learning_rate": 4.995030038851342e-06, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004822015180252492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017544238246046007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011398454243317247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3730, + "weighted_flow_matching_loss": 0.01669842703267932, + "weighted_rkd_loss": 0.0056992271216586236 + }, + { + "epoch": 6.275167785234899, + "flow_matching_loss": 0.01423853449523449, + "grad_norm": 0.15029947459697723, + "learning_rate": 4.9213333980415e-06, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005033561633899808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019928340218029915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01205995730124414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3740, + "weighted_flow_matching_loss": 0.01423853449523449, + "weighted_rkd_loss": 0.00602997865062207 + }, + { + "epoch": 6.291946308724833, + "flow_matching_loss": 0.016005691885948182, + "grad_norm": 0.1524253487586975, + "learning_rate": 4.848077630759175e-06, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004667837568558752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002215351990889758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011551027372479438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3750, + "weighted_flow_matching_loss": 0.016005691885948182, + "weighted_rkd_loss": 0.005775513686239719 + }, + { + "epoch": 6.308724832214765, + "flow_matching_loss": 0.017472335230559112, + "grad_norm": 0.27056488394737244, + "learning_rate": 4.775265941448951e-06, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005071670259349048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018874454370234161, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012030785949900747, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3760, + "weighted_flow_matching_loss": 0.017472335230559112, + "weighted_rkd_loss": 0.0060153929749503735 + }, + { + "epoch": 6.325503355704698, + "flow_matching_loss": 0.016118133440613747, + "grad_norm": 0.14759497344493866, + "learning_rate": 4.702901515129989e-06, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004513580771163106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019892573938705025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011016418971121312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3770, + "weighted_flow_matching_loss": 0.016118133440613747, + "weighted_rkd_loss": 0.005508209485560656 + }, + { + "epoch": 6.342281879194631, + "flow_matching_loss": 0.015037001110613346, + "grad_norm": 0.2157750427722931, + "learning_rate": 4.630987517256734e-06, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004265101510100067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001737450179643929, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010267653083428741, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3780, + "weighted_flow_matching_loss": 0.015037001110613346, + "weighted_rkd_loss": 0.005133826541714371 + }, + { + "epoch": 6.359060402684563, + "flow_matching_loss": 0.015858409553766252, + "grad_norm": 0.1426669806241989, + "learning_rate": 4.559527093580412e-06, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00464945612475276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019282410736195742, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011227153288200498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3790, + "weighted_flow_matching_loss": 0.015858409553766252, + "weighted_rkd_loss": 0.005613576644100249 + }, + { + "epoch": 6.375838926174497, + "flow_matching_loss": 0.015834339428693056, + "grad_norm": 0.11701840162277222, + "learning_rate": 4.488523370011456e-06, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038423003163188694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015366179984994232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00922121866606176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3800, + "weighted_flow_matching_loss": 0.015834339428693056, + "weighted_rkd_loss": 0.00461060933303088 + }, + { + "epoch": 6.39261744966443, + "flow_matching_loss": 0.014915316551923751, + "grad_norm": 0.12409483641386032, + "learning_rate": 4.417979452482766e-06, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003954608063213527, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016807083855383098, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009589924523606897, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3810, + "weighted_flow_matching_loss": 0.014915316551923751, + "weighted_rkd_loss": 0.0047949622618034485 + }, + { + "epoch": 6.409395973154362, + "flow_matching_loss": 0.014157744124531746, + "grad_norm": 0.11032804846763611, + "learning_rate": 4.347898426813811e-06, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004619434243068099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021324063884094356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011371274897828697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3820, + "weighted_flow_matching_loss": 0.014157744124531746, + "weighted_rkd_loss": 0.005685637448914349 + }, + { + "epoch": 6.426174496644295, + "flow_matching_loss": 0.014956588950008154, + "grad_norm": 0.14672960340976715, + "learning_rate": 4.278283358575694e-06, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004830200923606753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020344901247881355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011694891983643175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3830, + "weighted_flow_matching_loss": 0.014956588950008154, + "weighted_rkd_loss": 0.005847445991821587 + }, + { + "epoch": 6.442953020134228, + "flow_matching_loss": 0.014917949866503476, + "grad_norm": 0.14469048380851746, + "learning_rate": 4.2091372929570145e-06, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004648088780231774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002000366896390915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011296544317156076, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3840, + "weighted_flow_matching_loss": 0.014917949866503476, + "weighted_rkd_loss": 0.005648272158578038 + }, + { + "epoch": 6.459731543624161, + "flow_matching_loss": 0.01866527069360018, + "grad_norm": 0.17728039622306824, + "learning_rate": 4.1404632546306934e-06, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0048375848913565275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002021892520133406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011697062104940415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3850, + "weighted_flow_matching_loss": 0.01866527069360018, + "weighted_rkd_loss": 0.0058485310524702076 + }, + { + "epoch": 6.476510067114094, + "flow_matching_loss": 0.014754966367036104, + "grad_norm": 0.21448470652103424, + "learning_rate": 4.0722642476216286e-06, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00443953056819737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017713158624246717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010650376975536346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3860, + "weighted_flow_matching_loss": 0.014754966367036104, + "weighted_rkd_loss": 0.005325188487768173 + }, + { + "epoch": 6.493288590604027, + "flow_matching_loss": 0.01667445064522326, + "grad_norm": 0.14577926695346832, + "learning_rate": 4.00454325517533e-06, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004969216231256723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019304315675981343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011868863925337791, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3870, + "weighted_flow_matching_loss": 0.01667445064522326, + "weighted_rkd_loss": 0.0059344319626688955 + }, + { + "epoch": 6.510067114093959, + "flow_matching_loss": 0.016786788031458853, + "grad_norm": 0.1949937641620636, + "learning_rate": 3.937303239627393e-06, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004295885050669313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018998178886249662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010491588059812785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3880, + "weighted_flow_matching_loss": 0.016786788031458853, + "weighted_rkd_loss": 0.005245794029906392 + }, + { + "epoch": 6.526845637583893, + "flow_matching_loss": 0.01631803922355175, + "grad_norm": 0.21952104568481445, + "learning_rate": 3.870547142273918e-06, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004865309922024608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001927953865379095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011658573616296053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3890, + "weighted_flow_matching_loss": 0.01631803922355175, + "weighted_rkd_loss": 0.005829286808148026 + }, + { + "epoch": 6.543624161073826, + "flow_matching_loss": 0.016452026460319756, + "grad_norm": 0.196889266371727, + "learning_rate": 3.8042778832428736e-06, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004349646950140595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019458610797300935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010645155049860477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3900, + "weighted_flow_matching_loss": 0.016452026460319756, + "weighted_rkd_loss": 0.005322577524930238 + }, + { + "epoch": 6.560402684563758, + "flow_matching_loss": 0.019432486873120068, + "grad_norm": 0.13456007838249207, + "learning_rate": 3.7384983613663195e-06, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004305106052197516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016713534423615783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010281565645709635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3910, + "weighted_flow_matching_loss": 0.019432486873120068, + "weighted_rkd_loss": 0.005140782822854817 + }, + { + "epoch": 6.577181208053691, + "flow_matching_loss": 0.01630993653088808, + "grad_norm": 0.14141227304935455, + "learning_rate": 3.6732114540536503e-06, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041214486351236704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001664979092311114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009907876374199986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3920, + "weighted_flow_matching_loss": 0.01630993653088808, + "weighted_rkd_loss": 0.004953938187099993 + }, + { + "epoch": 6.593959731543624, + "flow_matching_loss": 0.016547040734440088, + "grad_norm": 0.178508922457695, + "learning_rate": 3.608420017165683e-06, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038512055296450855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015659712022170424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009268382238224149, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3930, + "weighted_flow_matching_loss": 0.016547040734440088, + "weighted_rkd_loss": 0.0046341911191120745 + }, + { + "epoch": 6.610738255033557, + "flow_matching_loss": 0.01753809815272689, + "grad_norm": 0.17572665214538574, + "learning_rate": 3.5441268848897664e-06, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004356299003120512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018225304782390594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010535128554329276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3940, + "weighted_flow_matching_loss": 0.01753809815272689, + "weighted_rkd_loss": 0.005267564277164638 + }, + { + "epoch": 6.62751677852349, + "flow_matching_loss": 0.015104027651250362, + "grad_norm": 0.17830194532871246, + "learning_rate": 3.480334869615792e-06, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004359930544160307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016910021542571485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010410863300785422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3950, + "weighted_flow_matching_loss": 0.015104027651250362, + "weighted_rkd_loss": 0.005205431650392711 + }, + { + "epoch": 6.644295302013423, + "flow_matching_loss": 0.015497284568846225, + "grad_norm": 0.15404456853866577, + "learning_rate": 3.417046761813155e-06, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004788657440803945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018028156249783934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011380130471661688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3960, + "weighted_flow_matching_loss": 0.015497284568846225, + "weighted_rkd_loss": 0.005690065235830844 + }, + { + "epoch": 6.6610738255033555, + "flow_matching_loss": 0.015348309185355902, + "grad_norm": 0.1278257668018341, + "learning_rate": 3.3542653299087263e-06, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004161874623969197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001591894228477031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009915643278509379, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3970, + "weighted_flow_matching_loss": 0.015348309185355902, + "weighted_rkd_loss": 0.004957821639254689 + }, + { + "epoch": 6.677852348993289, + "flow_matching_loss": 0.012799481302499771, + "grad_norm": 0.11342970281839371, + "learning_rate": 3.2919933201657117e-06, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004698016052134335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017846812726929784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011180713446810842, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3980, + "weighted_flow_matching_loss": 0.012799481302499771, + "weighted_rkd_loss": 0.005590356723405421 + }, + { + "epoch": 6.694630872483222, + "flow_matching_loss": 0.016074532829225063, + "grad_norm": 0.1375473588705063, + "learning_rate": 3.2302334565635584e-06, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004121939861215651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001517608918948099, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009761488577350975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3990, + "weighted_flow_matching_loss": 0.016074532829225063, + "weighted_rkd_loss": 0.004880744288675487 + }, + { + "epoch": 6.7114093959731544, + "flow_matching_loss": 0.013632395770400763, + "grad_norm": 0.13031578063964844, + "learning_rate": 3.1689884406787674e-06, + "loss": 0.019, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004645679029636085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015164404409006237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010807798663154244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4000, + "weighted_flow_matching_loss": 0.013632395770400763, + "weighted_rkd_loss": 0.005403899331577122 + }, + { + "epoch": 6.728187919463087, + "flow_matching_loss": 0.015095925237983466, + "grad_norm": 0.16176630556583405, + "learning_rate": 3.1082609515667424e-06, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0054400034714490175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021193060209043326, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012999312952160835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4010, + "weighted_flow_matching_loss": 0.015095925237983466, + "weighted_rkd_loss": 0.006499656476080418 + }, + { + "epoch": 6.74496644295302, + "flow_matching_loss": 0.01631878176704049, + "grad_norm": 0.1447662115097046, + "learning_rate": 3.048053645644594e-06, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005066341208294034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020264827413484453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012159165181219577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4020, + "weighted_flow_matching_loss": 0.01631878176704049, + "weighted_rkd_loss": 0.0060795825906097885 + }, + { + "epoch": 6.7617449664429525, + "flow_matching_loss": 0.01471923552453518, + "grad_norm": 0.17339105904102325, + "learning_rate": 2.988369156574916e-06, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004106980562210083, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014705379900988192, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00968449916690588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4030, + "weighted_flow_matching_loss": 0.01471923552453518, + "weighted_rkd_loss": 0.00484224958345294 + }, + { + "epoch": 6.778523489932886, + "flow_matching_loss": 0.01501200720667839, + "grad_norm": 0.13662545382976532, + "learning_rate": 2.9292100951506225e-06, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036912807961925866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015375688206404447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008920130459591747, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4040, + "weighted_flow_matching_loss": 0.01501200720667839, + "weighted_rkd_loss": 0.004460065229795873 + }, + { + "epoch": 6.795302013422819, + "flow_matching_loss": 0.016395939514040947, + "grad_norm": 0.16309326887130737, + "learning_rate": 2.8705790491807033e-06, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00422157181892544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016458123573102057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010088955983519554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4050, + "weighted_flow_matching_loss": 0.016395939514040947, + "weighted_rkd_loss": 0.005044477991759777 + }, + { + "epoch": 6.8120805369127515, + "flow_matching_loss": 0.014857204258441925, + "grad_norm": 0.19027967751026154, + "learning_rate": 2.8124785833770433e-06, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00493489746004343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019772169762291014, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011847011931240558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4060, + "weighted_flow_matching_loss": 0.014857204258441925, + "weighted_rkd_loss": 0.005923505965620279 + }, + { + "epoch": 6.828859060402684, + "flow_matching_loss": 0.01854044497013092, + "grad_norm": 0.12597495317459106, + "learning_rate": 2.754911239242241e-06, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004213976627215743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018678706022910775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010295823868364095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4070, + "weighted_flow_matching_loss": 0.01854044497013092, + "weighted_rkd_loss": 0.0051479119341820475 + }, + { + "epoch": 6.845637583892618, + "flow_matching_loss": 0.014754861127585173, + "grad_norm": 0.11598055809736252, + "learning_rate": 2.6978795349584204e-06, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033723325235769154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013575545453932136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008102219691500068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4080, + "weighted_flow_matching_loss": 0.014754861127585173, + "weighted_rkd_loss": 0.004051109845750034 + }, + { + "epoch": 6.8624161073825505, + "flow_matching_loss": 0.016834757197648285, + "grad_norm": 0.16511546075344086, + "learning_rate": 2.641385965277088e-06, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039168634917587045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017385203973390162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009572247229516506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4090, + "weighted_flow_matching_loss": 0.016834757197648285, + "weighted_rkd_loss": 0.004786123614758253 + }, + { + "epoch": 6.879194630872483, + "flow_matching_loss": 0.015211101062595845, + "grad_norm": 0.13418087363243103, + "learning_rate": 2.5854330014099923e-06, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004090139199979604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001788419799413532, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00996869821101427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4100, + "weighted_flow_matching_loss": 0.015211101062595845, + "weighted_rkd_loss": 0.004984349105507135 + }, + { + "epoch": 6.895973154362416, + "flow_matching_loss": 0.015671815164387225, + "grad_norm": 0.1589498072862625, + "learning_rate": 2.5300230909210298e-06, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004447044571861625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016958691005129368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010589958261698484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4110, + "weighted_flow_matching_loss": 0.015671815164387225, + "weighted_rkd_loss": 0.005294979130849242 + }, + { + "epoch": 6.912751677852349, + "flow_matching_loss": 0.014125759527087212, + "grad_norm": 0.1285037398338318, + "learning_rate": 2.4751586576191983e-06, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003744475147686899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014271572756115348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008916107751429082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4120, + "weighted_flow_matching_loss": 0.014125759527087212, + "weighted_rkd_loss": 0.004458053875714541 + }, + { + "epoch": 6.929530201342282, + "flow_matching_loss": 0.012916694115847348, + "grad_norm": 0.14929188787937164, + "learning_rate": 2.420842101452538e-06, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004445642535574734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001718758384231478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010610043350607157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4130, + "weighted_flow_matching_loss": 0.012916694115847348, + "weighted_rkd_loss": 0.005305021675303579 + }, + { + "epoch": 6.946308724832215, + "flow_matching_loss": 0.013612718507647514, + "grad_norm": 0.14932781457901, + "learning_rate": 2.3670757984031816e-06, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003692518384195864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014336197986267508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008818656718358398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4140, + "weighted_flow_matching_loss": 0.013612718507647514, + "weighted_rkd_loss": 0.004409328359179199 + }, + { + "epoch": 6.9630872483221475, + "flow_matching_loss": 0.013868955336511135, + "grad_norm": 0.1929025948047638, + "learning_rate": 2.313862100383404e-06, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004685251065529883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018140813626814634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01118458341807127, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4150, + "weighted_flow_matching_loss": 0.013868955336511135, + "weighted_rkd_loss": 0.005592291709035635 + }, + { + "epoch": 6.97986577181208, + "flow_matching_loss": 0.012972024455666541, + "grad_norm": 0.11274044960737228, + "learning_rate": 2.2612033351327406e-06, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003931707842275501, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015600805985741318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009423496341332794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4160, + "weighted_flow_matching_loss": 0.012972024455666541, + "weighted_rkd_loss": 0.004711748170666397 + }, + { + "epoch": 6.996644295302014, + "flow_matching_loss": 0.014395527262240648, + "grad_norm": 0.15545383095741272, + "learning_rate": 2.2091018061161788e-06, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004139221063815058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001665854500606656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009944296535104513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4170, + "weighted_flow_matching_loss": 0.014395527262240648, + "weighted_rkd_loss": 0.004972148267552256 + }, + { + "epoch": 7.0134228187919465, + "flow_matching_loss": 0.01504547274671495, + "grad_norm": 0.19873428344726562, + "learning_rate": 2.15755979242338e-06, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00428015049546957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001995652518235147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010555953532457352, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4180, + "weighted_flow_matching_loss": 0.01504547274671495, + "weighted_rkd_loss": 0.005277976766228676 + }, + { + "epoch": 7.030201342281879, + "flow_matching_loss": 0.016630757320672273, + "grad_norm": 0.12612350285053253, + "learning_rate": 2.106579548669e-06, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004597616638056934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016638765577226878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010859109880402684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4190, + "weighted_flow_matching_loss": 0.016630757320672273, + "weighted_rkd_loss": 0.005429554940201342 + }, + { + "epoch": 7.046979865771812, + "flow_matching_loss": 0.013459595013409853, + "grad_norm": 0.15308141708374023, + "learning_rate": 2.0561633048940477e-06, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040692856186069545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018304155499208718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009968986781314016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4200, + "weighted_flow_matching_loss": 0.013459595013409853, + "weighted_rkd_loss": 0.004984493390657008 + }, + { + "epoch": 7.063758389261745, + "flow_matching_loss": 0.016229463275521994, + "grad_norm": 0.15950238704681396, + "learning_rate": 2.006313266468355e-06, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004591330466791987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017486129305325448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010931274108588695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4210, + "weighted_flow_matching_loss": 0.016229463275521994, + "weighted_rkd_loss": 0.005465637054294347 + }, + { + "epoch": 7.080536912751678, + "flow_matching_loss": 0.0157932304777205, + "grad_norm": 0.1525679975748062, + "learning_rate": 1.957031613994097e-06, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038325243163853883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015466227720025926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00921167149208486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4220, + "weighted_flow_matching_loss": 0.0157932304777205, + "weighted_rkd_loss": 0.00460583574604243 + }, + { + "epoch": 7.097315436241611, + "flow_matching_loss": 0.016353191807866096, + "grad_norm": 0.1053260862827301, + "learning_rate": 1.9083205032103995e-06, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004086917964741588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016238337906543166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009797669760882855, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4230, + "weighted_flow_matching_loss": 0.016353191807866096, + "weighted_rkd_loss": 0.004898834880441428 + }, + { + "epoch": 7.114093959731544, + "flow_matching_loss": 0.01574868755415082, + "grad_norm": 0.1536342352628708, + "learning_rate": 1.8601820648990536e-06, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035363011527806522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014526980870869012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008525300398468971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4240, + "weighted_flow_matching_loss": 0.01574868755415082, + "weighted_rkd_loss": 0.0042626501992344855 + }, + { + "epoch": 7.130872483221476, + "flow_matching_loss": 0.014412057958543301, + "grad_norm": 0.13013885915279388, + "learning_rate": 1.8126184047912936e-06, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003889480617363006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015578919497784226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009336853120476007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4250, + "weighted_flow_matching_loss": 0.014412057958543301, + "weighted_rkd_loss": 0.004668426560238003 + }, + { + "epoch": 7.14765100671141, + "flow_matching_loss": 0.014226209558546542, + "grad_norm": 0.11760559678077698, + "learning_rate": 1.7656316034757024e-06, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038801743183285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018998766783624888, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009660225315019488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4260, + "weighted_flow_matching_loss": 0.014226209558546542, + "weighted_rkd_loss": 0.004830112657509744 + }, + { + "epoch": 7.1644295302013425, + "flow_matching_loss": 0.016309910733252764, + "grad_norm": 0.17850618064403534, + "learning_rate": 1.7192237163071816e-06, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036488832789473235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001311828033067286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008609594684094191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4270, + "weighted_flow_matching_loss": 0.016309910733252764, + "weighted_rkd_loss": 0.0043047973420470955 + }, + { + "epoch": 7.181208053691275, + "flow_matching_loss": 0.016244333423674106, + "grad_norm": 0.13397976756095886, + "learning_rate": 1.6733967733170573e-06, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004331002850085497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017532274941913784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010415233299136161, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4280, + "weighted_flow_matching_loss": 0.016244333423674106, + "weighted_rkd_loss": 0.005207616649568081 + }, + { + "epoch": 7.197986577181208, + "flow_matching_loss": 0.013507804181426764, + "grad_norm": 0.1460564285516739, + "learning_rate": 1.628152779124279e-06, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003937823371961713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016270102758426219, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009502657083794474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4290, + "weighted_flow_matching_loss": 0.013507804181426764, + "weighted_rkd_loss": 0.004751328541897237 + }, + { + "epoch": 7.214765100671141, + "flow_matching_loss": 0.014902317523956298, + "grad_norm": 0.12618981301784515, + "learning_rate": 1.583493712847715e-06, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037965486524626614, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015084690239746124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009101566206663848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4300, + "weighted_flow_matching_loss": 0.014902317523956298, + "weighted_rkd_loss": 0.004550783103331924 + }, + { + "epoch": 7.231543624161074, + "flow_matching_loss": 0.016011614352464676, + "grad_norm": 0.14825978875160217, + "learning_rate": 1.5394215280196084e-06, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037238317308947444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015631090383976698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009010772593319416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4310, + "weighted_flow_matching_loss": 0.016011614352464676, + "weighted_rkd_loss": 0.004505386296659708 + }, + { + "epoch": 7.248322147651007, + "flow_matching_loss": 0.015480771102011204, + "grad_norm": 0.12888647615909576, + "learning_rate": 1.4959381525000904e-06, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004318940313532948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016981088323518635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01033598962239921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4320, + "weighted_flow_matching_loss": 0.015480771102011204, + "weighted_rkd_loss": 0.005167994811199605 + }, + { + "epoch": 7.26510067114094, + "flow_matching_loss": 0.01660889098420739, + "grad_norm": 0.12845715880393982, + "learning_rate": 1.453045488392879e-06, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003837773168925196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013941763783805073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009069722797721624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4330, + "weighted_flow_matching_loss": 0.01660889098420739, + "weighted_rkd_loss": 0.004534861398860812 + }, + { + "epoch": 7.281879194630872, + "flow_matching_loss": 0.01536073824390769, + "grad_norm": 0.1266445517539978, + "learning_rate": 1.4107454119620484e-06, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004038789356127381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014731727307662368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009550751419737935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4340, + "weighted_flow_matching_loss": 0.01536073824390769, + "weighted_rkd_loss": 0.0047753757098689675 + }, + { + "epoch": 7.298657718120805, + "flow_matching_loss": 0.015487385308369994, + "grad_norm": 0.1259327232837677, + "learning_rate": 1.3690397735499778e-06, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004094103630632162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016427026013843716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009830909874290228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4350, + "weighted_flow_matching_loss": 0.015487385308369994, + "weighted_rkd_loss": 0.004915454937145114 + }, + { + "epoch": 7.315436241610739, + "flow_matching_loss": 0.016292981710284947, + "grad_norm": 0.208975687623024, + "learning_rate": 1.327930397496403e-06, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038555726176127793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014661048422567546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009177249949425459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4360, + "weighted_flow_matching_loss": 0.016292981710284947, + "weighted_rkd_loss": 0.004588624974712729 + }, + { + "epoch": 7.332214765100671, + "flow_matching_loss": 0.013602617662400007, + "grad_norm": 0.11648474633693695, + "learning_rate": 1.287419082058604e-06, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004054751852527261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015656212577596306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009675125079229473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4370, + "weighted_flow_matching_loss": 0.013602617662400007, + "weighted_rkd_loss": 0.004837562539614737 + }, + { + "epoch": 7.348993288590604, + "flow_matching_loss": 0.015651093889027835, + "grad_norm": 0.12388869374990463, + "learning_rate": 1.2475075993327589e-06, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003548008017241955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001444374053971842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008540389966219664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4380, + "weighted_flow_matching_loss": 0.015651093889027835, + "weighted_rkd_loss": 0.004270194983109832 + }, + { + "epoch": 7.365771812080537, + "flow_matching_loss": 0.013896638620644807, + "grad_norm": 0.13266593217849731, + "learning_rate": 1.208197695176413e-06, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004024438234046102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016310391831211746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00967991566285491, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4390, + "weighted_flow_matching_loss": 0.013896638620644807, + "weighted_rkd_loss": 0.004839957831427455 + }, + { + "epoch": 7.382550335570469, + "flow_matching_loss": 0.013670449983328581, + "grad_norm": 0.17349274456501007, + "learning_rate": 1.1694910891321259e-06, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004115662141703069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001552202249877155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009783526510000229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4400, + "weighted_flow_matching_loss": 0.013670449983328581, + "weighted_rkd_loss": 0.004891763255000114 + }, + { + "epoch": 7.399328859060403, + "flow_matching_loss": 0.014937188103795051, + "grad_norm": 0.23631052672863007, + "learning_rate": 1.1313894743522363e-06, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042281194357201455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001866805658210069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010323044611141086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4410, + "weighted_flow_matching_loss": 0.014937188103795051, + "weighted_rkd_loss": 0.005161522305570543 + }, + { + "epoch": 7.416107382550336, + "flow_matching_loss": 0.015254345163702965, + "grad_norm": 0.12517230212688446, + "learning_rate": 1.0938945175248016e-06, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003667901479639113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001594919990748167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008930722950026393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4420, + "weighted_flow_matching_loss": 0.015254345163702965, + "weighted_rkd_loss": 0.0044653614750131965 + }, + { + "epoch": 7.432885906040268, + "flow_matching_loss": 0.01722643505781889, + "grad_norm": 0.1511530727148056, + "learning_rate": 1.0570078588007086e-06, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004720710683614016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017709346022456885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011212356062605977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4430, + "weighted_flow_matching_loss": 0.01722643505781889, + "weighted_rkd_loss": 0.005606178031302989 + }, + { + "epoch": 7.449664429530201, + "flow_matching_loss": 0.014523876365274191, + "grad_norm": 0.2035868763923645, + "learning_rate": 1.0207311117219009e-06, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004571941145695746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018188763991929592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01096275863237679, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4440, + "weighted_flow_matching_loss": 0.014523876365274191, + "weighted_rkd_loss": 0.005481379316188395 + }, + { + "epoch": 7.466442953020135, + "flow_matching_loss": 0.013871367648243904, + "grad_norm": 0.1734429895877838, + "learning_rate": 9.850658631508198e-07, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003869479591958225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015727715042885392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009311730740591884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4450, + "weighted_flow_matching_loss": 0.013871367648243904, + "weighted_rkd_loss": 0.004655865370295942 + }, + { + "epoch": 7.483221476510067, + "flow_matching_loss": 0.0138609291985631, + "grad_norm": 0.16242648661136627, + "learning_rate": 9.500136732009751e-07, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004720749869011343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016571854008361696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01109868511557579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4460, + "weighted_flow_matching_loss": 0.0138609291985631, + "weighted_rkd_loss": 0.005549342557787895 + }, + { + "epoch": 7.5, + "flow_matching_loss": 0.013707988150417804, + "grad_norm": 0.12876121699810028, + "learning_rate": 9.155760751687054e-07, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003897508908994496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001529338606633246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009324356354773045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4470, + "weighted_flow_matching_loss": 0.013707988150417804, + "weighted_rkd_loss": 0.0046621781773865225 + }, + { + "epoch": 7.516778523489933, + "flow_matching_loss": 0.014499800279736519, + "grad_norm": 0.16918422281742096, + "learning_rate": 8.817545754661155e-07, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004376136627979576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016412930912338198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010393566358834505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4480, + "weighted_flow_matching_loss": 0.014499800279736519, + "weighted_rkd_loss": 0.005196783179417253 + }, + { + "epoch": 7.533557046979865, + "flow_matching_loss": 0.014267795532941819, + "grad_norm": 0.16523607075214386, + "learning_rate": 8.485506535551668e-07, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004032965656369925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016446026973426342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009710533916950226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4490, + "weighted_flow_matching_loss": 0.014267795532941819, + "weighted_rkd_loss": 0.004855266958475113 + }, + { + "epoch": 7.550335570469799, + "flow_matching_loss": 0.01365661434829235, + "grad_norm": 0.1421973556280136, + "learning_rate": 8.159657618829685e-07, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037587011931464074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014382767374627292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008955679135397077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4500, + "weighted_flow_matching_loss": 0.01365661434829235, + "weighted_rkd_loss": 0.004477839567698538 + }, + { + "epoch": 7.567114093959732, + "flow_matching_loss": 0.015886571630835532, + "grad_norm": 0.14739073812961578, + "learning_rate": 7.840013258182471e-07, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040415175259113315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015700169489718974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009653051989153028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4510, + "weighted_flow_matching_loss": 0.015886571630835532, + "weighted_rkd_loss": 0.004826525994576514 + }, + { + "epoch": 7.583892617449664, + "flow_matching_loss": 0.015283238515257836, + "grad_norm": 0.18063873052597046, + "learning_rate": 7.526587435889809e-07, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004281807481311262, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019427814171649516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010506396507844329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4520, + "weighted_flow_matching_loss": 0.015283238515257836, + "weighted_rkd_loss": 0.005253198253922164 + }, + { + "epoch": 7.600671140939597, + "flow_matching_loss": 0.015953387320041656, + "grad_norm": 0.1318785548210144, + "learning_rate": 7.219393862212593e-07, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004065710003487766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001621024962514639, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009752445016056299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4530, + "weighted_flow_matching_loss": 0.015953387320041656, + "weighted_rkd_loss": 0.004876222508028149 + }, + { + "epoch": 7.617449664429531, + "flow_matching_loss": 0.015033760201185941, + "grad_norm": 0.18385183811187744, + "learning_rate": 6.918445974792819e-07, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004502101079560816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018878166214562953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010892018815502525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4540, + "weighted_flow_matching_loss": 0.015033760201185941, + "weighted_rkd_loss": 0.005446009407751262 + }, + { + "epoch": 7.634228187919463, + "flow_matching_loss": 0.015050337370485067, + "grad_norm": 0.1594400852918625, + "learning_rate": 6.6237569380661e-07, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004070810438133776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00172933719586581, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009870958141982555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4550, + "weighted_flow_matching_loss": 0.015050337370485067, + "weighted_rkd_loss": 0.0049354790709912775 + }, + { + "epoch": 7.651006711409396, + "flow_matching_loss": 0.014827259257435799, + "grad_norm": 0.21137267351150513, + "learning_rate": 6.335339642685495e-07, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043532455572858455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001815312390681356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010521803423762322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4560, + "weighted_flow_matching_loss": 0.014827259257435799, + "weighted_rkd_loss": 0.005260901711881161 + }, + { + "epoch": 7.667785234899329, + "flow_matching_loss": 0.014157627988606692, + "grad_norm": 0.15432624518871307, + "learning_rate": 6.05320670495787e-07, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003650864795781672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015314153046347202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008833145024254918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4570, + "weighted_flow_matching_loss": 0.014157627988606692, + "weighted_rkd_loss": 0.004416572512127459 + }, + { + "epoch": 7.684563758389261, + "flow_matching_loss": 0.013985906587913632, + "grad_norm": 0.12158872187137604, + "learning_rate": 5.777370466291954e-07, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003960758331231773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015997558715753258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009521272592246532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4580, + "weighted_flow_matching_loss": 0.013985906587913632, + "weighted_rkd_loss": 0.004760636296123266 + }, + { + "epoch": 7.701342281879195, + "flow_matching_loss": 0.014527055993676185, + "grad_norm": 0.11905797570943832, + "learning_rate": 5.507842992658336e-07, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003300216724164784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001481162616983056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008081595972180366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4590, + "weighted_flow_matching_loss": 0.014527055993676185, + "weighted_rkd_loss": 0.004040797986090183 + }, + { + "epoch": 7.718120805369128, + "flow_matching_loss": 0.01651501702144742, + "grad_norm": 0.144264355301857, + "learning_rate": 5.24463607406192e-07, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004112446680665016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001520359597634524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009745252970606088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4600, + "weighted_flow_matching_loss": 0.01651501702144742, + "weighted_rkd_loss": 0.004872626485303044 + }, + { + "epoch": 7.73489932885906, + "flow_matching_loss": 0.013798393215984106, + "grad_norm": 0.18188291788101196, + "learning_rate": 4.987761224025944e-07, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004624183708801866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015590140363201499, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010807381616905332, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4610, + "weighted_flow_matching_loss": 0.013798393215984106, + "weighted_rkd_loss": 0.005403690808452666 + }, + { + "epoch": 7.751677852348993, + "flow_matching_loss": 0.018524532951414586, + "grad_norm": 0.12207607179880142, + "learning_rate": 4.737229679088556e-07, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004238227335736156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001529688062146306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010006142687052489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4620, + "weighted_flow_matching_loss": 0.018524532951414586, + "weighted_rkd_loss": 0.005003071343526244 + }, + { + "epoch": 7.768456375838926, + "flow_matching_loss": 0.01555617032572627, + "grad_norm": 0.2024114578962326, + "learning_rate": 4.493052398311115e-07, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00375926629640162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001470496621914208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008989029238000512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4630, + "weighted_flow_matching_loss": 0.01555617032572627, + "weighted_rkd_loss": 0.004494514619000256 + }, + { + "epoch": 7.785234899328859, + "flow_matching_loss": 0.015568674076348543, + "grad_norm": 0.14492373168468475, + "learning_rate": 4.2552400627989043e-07, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039431073004379865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001659181946888566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00954539654776454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4640, + "weighted_flow_matching_loss": 0.015568674076348543, + "weighted_rkd_loss": 0.00477269827388227 + }, + { + "epoch": 7.802013422818792, + "flow_matching_loss": 0.01527152806520462, + "grad_norm": 0.1488606482744217, + "learning_rate": 4.023803075233923e-07, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033758064033463596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001500186522025615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008251799317076803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4650, + "weighted_flow_matching_loss": 0.01527152806520462, + "weighted_rkd_loss": 0.0041258996585384015 + }, + { + "epoch": 7.818791946308725, + "flow_matching_loss": 0.015966273937374355, + "grad_norm": 0.12519210577011108, + "learning_rate": 3.7987515594196997e-07, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004246301972307265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016014743596315384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010094078304246069, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4660, + "weighted_flow_matching_loss": 0.015966273937374355, + "weighted_rkd_loss": 0.005047039152123034 + }, + { + "epoch": 7.8355704697986575, + "flow_matching_loss": 0.015654311794787646, + "grad_norm": 0.19098609685897827, + "learning_rate": 3.580095359838648e-07, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036975160473957658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014155545737594365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008810586575418711, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4670, + "weighted_flow_matching_loss": 0.015654311794787646, + "weighted_rkd_loss": 0.004405293287709356 + }, + { + "epoch": 7.85234899328859, + "flow_matching_loss": 0.015151944477111101, + "grad_norm": 0.131342351436615, + "learning_rate": 3.3678440412212576e-07, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00355694442987442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013498203014023602, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008463709149509668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4680, + "weighted_flow_matching_loss": 0.015151944477111101, + "weighted_rkd_loss": 0.004231854574754834 + }, + { + "epoch": 7.869127516778524, + "flow_matching_loss": 0.015845899749547244, + "grad_norm": 0.17855486273765564, + "learning_rate": 3.162006888127833e-07, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004481368698179722, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001695212651975453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010657949931919574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4690, + "weighted_flow_matching_loss": 0.015845899749547244, + "weighted_rkd_loss": 0.005328974965959787 + }, + { + "epoch": 7.885906040268456, + "flow_matching_loss": 0.017545492481440307, + "grad_norm": 0.1288793385028839, + "learning_rate": 2.9625929045422327e-07, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004457078292034566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017358578566927462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010650014318525792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4700, + "weighted_flow_matching_loss": 0.017545492481440307, + "weighted_rkd_loss": 0.005325007159262896 + }, + { + "epoch": 7.902684563758389, + "flow_matching_loss": 0.01548297619447112, + "grad_norm": 0.15775883197784424, + "learning_rate": 2.7696108134781186e-07, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004006052413024009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015338835888542234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009545988403260708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4710, + "weighted_flow_matching_loss": 0.01548297619447112, + "weighted_rkd_loss": 0.004772994201630354 + }, + { + "epoch": 7.919463087248322, + "flow_matching_loss": 0.017314035911113024, + "grad_norm": 0.17367108166217804, + "learning_rate": 2.583069056597376e-07, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00400707998778671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016400735126808285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009654233604669571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4720, + "weighted_flow_matching_loss": 0.017314035911113024, + "weighted_rkd_loss": 0.004827116802334786 + }, + { + "epoch": 7.936241610738255, + "flow_matching_loss": 0.013726966083049774, + "grad_norm": 0.1574770212173462, + "learning_rate": 2.4029757938407116e-07, + "loss": 0.0193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004623283655382693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019199330708943307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011166500393301249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4730, + "weighted_flow_matching_loss": 0.013726966083049774, + "weighted_rkd_loss": 0.005583250196650624 + }, + { + "epoch": 7.953020134228188, + "flow_matching_loss": 0.01642540153115988, + "grad_norm": 0.20747427642345428, + "learning_rate": 2.229338903070871e-07, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003777121054008603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015185043797828258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0090727464761585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4740, + "weighted_flow_matching_loss": 0.01642540153115988, + "weighted_rkd_loss": 0.00453637323807925 + }, + { + "epoch": 7.969798657718121, + "flow_matching_loss": 0.01649800632148981, + "grad_norm": 0.23679505288600922, + "learning_rate": 2.0621659797279645e-07, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004173471452668309, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016695614729542285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010016504442319274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4750, + "weighted_flow_matching_loss": 0.01649800632148981, + "weighted_rkd_loss": 0.005008252221159637 + }, + { + "epoch": 7.9865771812080535, + "flow_matching_loss": 0.015903161372989416, + "grad_norm": 0.12486451864242554, + "learning_rate": 1.901464336497233e-07, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004392204666510224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001676009374205023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01046041864901781, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4760, + "weighted_flow_matching_loss": 0.015903161372989416, + "weighted_rkd_loss": 0.005230209324508905 + }, + { + "epoch": 8.003355704697986, + "flow_matching_loss": 0.014692566078156232, + "grad_norm": 0.12237241864204407, + "learning_rate": 1.7472410029891527e-07, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040526770986616615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014504189952276648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009555773250758648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4770, + "weighted_flow_matching_loss": 0.014692566078156232, + "weighted_rkd_loss": 0.004777886625379324 + }, + { + "epoch": 8.020134228187919, + "flow_matching_loss": 0.014108181092888116, + "grad_norm": 0.10870523005723953, + "learning_rate": 1.5995027254319173e-07, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040929106064140795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001453663141001016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009639484342187643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4780, + "weighted_flow_matching_loss": 0.014108181092888116, + "weighted_rkd_loss": 0.0048197421710938215 + }, + { + "epoch": 8.036912751677852, + "flow_matching_loss": 0.0141829714179039, + "grad_norm": 0.12840935587882996, + "learning_rate": 1.4582559663764372e-07, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034840189618989825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014725176617503167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008440555585548282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4790, + "weighted_flow_matching_loss": 0.0141829714179039, + "weighted_rkd_loss": 0.004220277792774141 + }, + { + "epoch": 8.053691275167786, + "flow_matching_loss": 0.016363996639847757, + "grad_norm": 0.13803021609783173, + "learning_rate": 1.323506904413485e-07, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003999671293422579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015343882550951093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009533730894327163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4800, + "weighted_flow_matching_loss": 0.016363996639847757, + "weighted_rkd_loss": 0.004766865447163582 + }, + { + "epoch": 8.070469798657719, + "flow_matching_loss": 0.014740627631545067, + "grad_norm": 0.15675628185272217, + "learning_rate": 1.195261433903594e-07, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036474255612120032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014107772847637534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008705628290772437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4810, + "weighted_flow_matching_loss": 0.014740627631545067, + "weighted_rkd_loss": 0.004352814145386219 + }, + { + "epoch": 8.087248322147651, + "flow_matching_loss": 0.015853640623390675, + "grad_norm": 0.21039828658103943, + "learning_rate": 1.0735251647190969e-07, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004786642431281507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017930870177224278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01136637181043625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4820, + "weighted_flow_matching_loss": 0.015853640623390675, + "weighted_rkd_loss": 0.005683185905218125 + }, + { + "epoch": 8.104026845637584, + "flow_matching_loss": 0.013481860840693116, + "grad_norm": 0.14539654552936554, + "learning_rate": 9.583034219987408e-08, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0032999050919897853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014778119686525315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008077622065320611, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4830, + "weighted_flow_matching_loss": 0.013481860840693116, + "weighted_rkd_loss": 0.004038811032660306 + }, + { + "epoch": 8.120805369127517, + "flow_matching_loss": 0.012944562174379826, + "grad_norm": 0.1711987853050232, + "learning_rate": 8.496012459148717e-08, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00421827076934278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00177860067342408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010215142229571938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4840, + "weighted_flow_matching_loss": 0.012944562174379826, + "weighted_rkd_loss": 0.005107571114785969 + }, + { + "epoch": 8.13758389261745, + "flow_matching_loss": 0.016942327935248615, + "grad_norm": 0.16165891289710999, + "learning_rate": 7.474233914527618e-08, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038894150173291566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016811634763143958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00945999352261424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4850, + "weighted_flow_matching_loss": 0.016942327935248615, + "weighted_rkd_loss": 0.00472999676130712 + }, + { + "epoch": 8.154362416107382, + "flow_matching_loss": 0.016499594505876302, + "grad_norm": 0.20303112268447876, + "learning_rate": 6.517743282028254e-08, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003583537461236119, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014558300492353738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008622905006632209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4860, + "weighted_flow_matching_loss": 0.016499594505876302, + "weighted_rkd_loss": 0.004311452503316104 + }, + { + "epoch": 8.171140939597315, + "flow_matching_loss": 0.01599147580564022, + "grad_norm": 0.17316727340221405, + "learning_rate": 5.6265824016492604e-08, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003953036200255155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017714102636091412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009677482675760985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4870, + "weighted_flow_matching_loss": 0.01599147580564022, + "weighted_rkd_loss": 0.004838741337880492 + }, + { + "epoch": 8.187919463087248, + "flow_matching_loss": 0.016761266998946665, + "grad_norm": 0.1150829866528511, + "learning_rate": 4.800790255654719e-08, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036187476478517057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013934143644291908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008630909631028772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4880, + "weighted_flow_matching_loss": 0.016761266998946665, + "weighted_rkd_loss": 0.004315454815514386 + }, + { + "epoch": 8.204697986577182, + "flow_matching_loss": 0.017060472443699837, + "grad_norm": 0.19914576411247253, + "learning_rate": 4.040402966867862e-08, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00388106107711792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015067332715261728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00926885548979044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4890, + "weighted_flow_matching_loss": 0.017060472443699837, + "weighted_rkd_loss": 0.00463442774489522 + }, + { + "epoch": 8.221476510067115, + "flow_matching_loss": 0.01452061990275979, + "grad_norm": 0.17426694929599762, + "learning_rate": 3.3454537970925014e-08, + "loss": 0.019, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003739560558460653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014982009073719383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008977322094142436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4900, + "weighted_flow_matching_loss": 0.01452061990275979, + "weighted_rkd_loss": 0.004488661047071218 + }, + { + "epoch": 8.238255033557047, + "flow_matching_loss": 0.01707121105864644, + "grad_norm": 0.15827062726020813, + "learning_rate": 2.715973145656192e-08, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036373403389006855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014872469764668494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008761927671730518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4910, + "weighted_flow_matching_loss": 0.01707121105864644, + "weighted_rkd_loss": 0.004380963835865259 + }, + { + "epoch": 8.25503355704698, + "flow_matching_loss": 0.013726630993187428, + "grad_norm": 0.13560862839221954, + "learning_rate": 2.151988548082129e-08, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038332486758008598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015756478009279818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009242145158350468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4920, + "weighted_flow_matching_loss": 0.013726630993187428, + "weighted_rkd_loss": 0.004621072579175234 + }, + { + "epoch": 8.271812080536913, + "flow_matching_loss": 0.015442668460309505, + "grad_norm": 0.20286814868450165, + "learning_rate": 1.6535246748837772e-08, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0048883697018027306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002030312130227685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011807051626965404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4930, + "weighted_flow_matching_loss": 0.015442668460309505, + "weighted_rkd_loss": 0.005903525813482702 + }, + { + "epoch": 8.288590604026846, + "flow_matching_loss": 0.015910115744918584, + "grad_norm": 0.1755928248167038, + "learning_rate": 1.2206033304855702e-08, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004257333232089877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016399717191234232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010154638253152371, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4940, + "weighted_flow_matching_loss": 0.015910115744918584, + "weighted_rkd_loss": 0.005077319126576186 + }, + { + "epoch": 8.305369127516778, + "flow_matching_loss": 0.017014620639383794, + "grad_norm": 0.21290400624275208, + "learning_rate": 8.532434522695031e-09, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004705283232033253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018529294757172465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011263495916500688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4950, + "weighted_flow_matching_loss": 0.017014620639383794, + "weighted_rkd_loss": 0.005631747958250344 + }, + { + "epoch": 8.322147651006711, + "flow_matching_loss": 0.015421977546066046, + "grad_norm": 0.139173686504364, + "learning_rate": 5.5146110974679805e-09, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003743798425421119, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014029679470695556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008890564925968647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4960, + "weighted_flow_matching_loss": 0.015421977546066046, + "weighted_rkd_loss": 0.004445282462984324 + }, + { + "epoch": 8.338926174496644, + "flow_matching_loss": 0.01395677267573774, + "grad_norm": 0.1474369466304779, + "learning_rate": 3.1526950385496467e-09, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004145475127734244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015348056796938181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009825755842030048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4970, + "weighted_flow_matching_loss": 0.01395677267573774, + "weighted_rkd_loss": 0.004912877921015024 + }, + { + "epoch": 8.355704697986576, + "flow_matching_loss": 0.014628880564123392, + "grad_norm": 0.1921273171901703, + "learning_rate": 1.4467896637959666e-09, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004135441291145981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001468336395919323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00973921879194677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4980, + "weighted_flow_matching_loss": 0.014628880564123392, + "weighted_rkd_loss": 0.004869609395973385 + }, + { + "epoch": 8.37248322147651, + "flow_matching_loss": 0.01582834655418992, + "grad_norm": 0.17792509496212006, + "learning_rate": 3.9696959503898733e-10, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004110996657982468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001513004262233153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009734997665509582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4990, + "weighted_flow_matching_loss": 0.01582834655418992, + "weighted_rkd_loss": 0.004867498832754791 + }, + { + "epoch": 8.389261744966444, + "flow_matching_loss": 0.015212259162217379, + "grad_norm": 0.18235495686531067, + "learning_rate": 3.280754806156416e-12, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036568080307915808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00138852980453521, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00870214574970305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5000, + "weighted_flow_matching_loss": 0.015212259162217379, + "weighted_rkd_loss": 0.004351072874851525 + }, + { + "epoch": 8.389261744966444, + "step": 5000, + "total_flos": 0.0, + "train_loss": 0.04190636519491672, + "train_runtime": 8839.5869, + "train_samples_per_second": 36.201, + "train_steps_per_second": 0.566 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": true + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/training_args.bin b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/training_args.bin new file mode 100644 index 0000000000000000000000000000000000000000..1708255eebf76386fa8cdc23a2c77b89fbdb59c1 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/training_args.bin @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd17c2d04ff4a5d8327c39f884e977693edc7fe00a90e05bf06f81b4ec7cce31 +size 6097 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/resolved_config.yaml b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/resolved_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..b7fc2a6c3587ecefb101945fc2fa6de80e4694c2 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/resolved_config.yaml @@ -0,0 +1,38 @@ +experiment_name: fewshot_39tasks_5shot_from_rkd_timewar_processingline_v2_seed1 +policy_type: groot_RKDv3 +base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3 +output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1 +dataset: + dataset_soup: fewshot_heldout_atomic39_human_5shot +training: + num_gpus: 1 + batch_size: 64 + seed: 42 + filter_key_seed: 1 + manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed1.json + save_total_limit: 4 +model: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null +phases: +- name: train_5shot_seed1_processingline_5k + max_steps: 5000 + save_steps: 0 + save_step_list: + - 1000 + - 3000 + - 5000 + trainable: + preset: processing_line_only + tune_llm: false + tune_visual: false + tune_projector: false + tune_diffusion_model: false + losses: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null +resolved_sweep: {} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/seed1.yaml b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/seed1.yaml new file mode 100644 index 0000000000000000000000000000000000000000..d22b9a30765b65d093f1bce06f2e0a91ffcda67e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/seed1.yaml @@ -0,0 +1,37 @@ +experiment_name: fewshot_39tasks_5shot_from_rkd_timewar_processingline_v2_seed1 +policy_type: groot_RKDv3 +base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3 +output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1 +dataset: + dataset_soup: fewshot_heldout_atomic39_human_5shot +training: + num_gpus: 1 + batch_size: 64 + seed: 42 + filter_key_seed: 1 + manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed1.json + save_total_limit: 4 +model: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null +phases: +- name: train_5shot_seed1_processingline_5k + max_steps: 5000 + save_steps: 0 + save_step_list: + - 1000 + - 3000 + - 5000 + trainable: + preset: processing_line_only + tune_llm: false + tune_visual: false + tune_projector: false + tune_diffusion_model: false + losses: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 0.5 + rkd_loss_weight_schedule: null diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..27700cac5d292d4d42897b3c2e80bdc58e55c60e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5719731235192254, + -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-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a3758b2dac9cab1274647a0a3e4647bb197ec7fc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6f702e66e76c9c2a5773af5ec6b03a2aabb911ab03f04ce0b6ef1d2a8dca66a +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..100d8700122e08f81285427a03b4b056e9f4008d --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:34f25efe5de5b12e6ea97ec0a00e7fdb1025b149a557be648383e27eb29f97d8 +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/optimizer.pt b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..c3f2c2854dcd2f9aae494a0f5cc984f53ad3e2a1 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d3685461bdab9850a47cdec0092c35683cac281554dbf8ed0e2eba9fc9d1e3e1 +size 1624159435 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..6d7e29bdacdd0a4c34925c9a4c55901f4afed294 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-1000/trainer_state.json @@ -0,0 +1,4634 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.675041876046901, + "eval_steps": 500, + "global_step": 1000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.01675041876046901, + "flow_matching_loss": 0.20356450378894805, + "grad_norm": 1.0490323305130005, + "learning_rate": 1.08e-06, + "loss": 0.4004, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1398262470960617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11395216882228851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.39360466301441194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.20356450378894805, + "weighted_rkd_loss": 0.19680233150720597 + }, + { + "epoch": 0.03350083752093802, + "flow_matching_loss": 0.21135056614875794, + "grad_norm": 0.8279632925987244, + "learning_rate": 2.28e-06, + "loss": 0.3916, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12995845153927804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10064354985952377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3605604529380798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.21135056614875794, + "weighted_rkd_loss": 0.1802802264690399 + }, + { + "epoch": 0.05025125628140704, + "flow_matching_loss": 0.19959198087453842, + "grad_norm": 0.8481504321098328, + "learning_rate": 3.48e-06, + "loss": 0.3868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13302332013845444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1083116926252842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.374358332157135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.19959198087453842, + "weighted_rkd_loss": 0.1871791660785675 + }, + { + "epoch": 0.06700167504187604, + "flow_matching_loss": 0.21401427686214447, + "grad_norm": 1.1905933618545532, + "learning_rate": 4.68e-06, + "loss": 0.3892, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12562137991189956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09915168732404708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3503944456577301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.21401427686214447, + "weighted_rkd_loss": 0.17519722282886505 + }, + { + "epoch": 0.08375209380234507, + "flow_matching_loss": 0.18060285598039627, + "grad_norm": 0.9187169075012207, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11310266405344009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08473661690950393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3109419494867325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.18060285598039627, + "weighted_rkd_loss": 0.15547097474336624 + }, + { + "epoch": 0.10050251256281408, + "flow_matching_loss": 0.1746109426021576, + "grad_norm": 0.9370708465576172, + "learning_rate": 7.08e-06, + "loss": 0.3298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11362355500459671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08309690579771996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.31034401059150696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1746109426021576, + "weighted_rkd_loss": 0.15517200529575348 + }, + { + "epoch": 0.11725293132328309, + "flow_matching_loss": 0.16165669709444047, + "grad_norm": 0.7368708848953247, + "learning_rate": 8.28e-06, + "loss": 0.3148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11314839944243431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07991818860173225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30621498823165894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.16165669709444047, + "weighted_rkd_loss": 0.15310749411582947 + }, + { + "epoch": 0.13400335008375208, + "flow_matching_loss": 0.15373115316033364, + "grad_norm": 0.8529441356658936, + "learning_rate": 9.48e-06, + "loss": 0.2947, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10446386188268661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0729650903493166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.28189281225204466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.15373115316033364, + "weighted_rkd_loss": 0.14094640612602233 + }, + { + "epoch": 0.1507537688442211, + "flow_matching_loss": 0.12661252617836, + "grad_norm": 0.7936683893203735, + "learning_rate": 1.068e-05, + "loss": 0.2582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09823484048247337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06663991399109363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2631095930933952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.12661252617836, + "weighted_rkd_loss": 0.1315547965466976 + }, + { + "epoch": 0.16750418760469013, + "flow_matching_loss": 0.10844202637672425, + "grad_norm": 0.6985727548599243, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.2325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09290647432208061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.062390904128551486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24820385426282882, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.10844202637672425, + "weighted_rkd_loss": 0.12410192713141441 + }, + { + "epoch": 0.18425460636515914, + "flow_matching_loss": 0.09378867372870445, + "grad_norm": 0.6144883036613464, + "learning_rate": 1.308e-05, + "loss": 0.2126, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08944455981254577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05871000774204731, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23759912848472595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09378867372870445, + "weighted_rkd_loss": 0.11879956424236297 + }, + { + "epoch": 0.20100502512562815, + "flow_matching_loss": 0.08636006340384483, + "grad_norm": 0.519714891910553, + "learning_rate": 1.428e-05, + "loss": 0.1946, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08254799991846085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0513349212706089, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21643092334270478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08636006340384483, + "weighted_rkd_loss": 0.10821546167135239 + }, + { + "epoch": 0.21775544388609716, + "flow_matching_loss": 0.07800828218460083, + "grad_norm": 0.6437940001487732, + "learning_rate": 1.548e-05, + "loss": 0.1787, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07631000094115734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.048831694200634955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20145169496536255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07800828218460083, + "weighted_rkd_loss": 0.10072584748268128 + }, + { + "epoch": 0.23450586264656617, + "flow_matching_loss": 0.08341701701283455, + "grad_norm": 0.6835018992424011, + "learning_rate": 1.6680000000000003e-05, + "loss": 0.1816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0760519951581955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04422654993832111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19633053839206696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08341701701283455, + "weighted_rkd_loss": 0.09816526919603348 + }, + { + "epoch": 0.25125628140703515, + "flow_matching_loss": 0.07501983791589736, + "grad_norm": 0.6060196161270142, + "learning_rate": 1.7879999999999998e-05, + "loss": 0.1644, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06893355883657933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04092012867331505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17878724485635758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07501983791589736, + "weighted_rkd_loss": 0.08939362242817879 + }, + { + "epoch": 0.26800670016750416, + "flow_matching_loss": 0.07575965747237205, + "grad_norm": 0.5369148254394531, + "learning_rate": 1.908e-05, + "loss": 0.1611, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06494721546769142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.040794138610363004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17068857252597808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07575965747237205, + "weighted_rkd_loss": 0.08534428626298904 + }, + { + "epoch": 0.2847571189279732, + "flow_matching_loss": 0.07186435237526893, + "grad_norm": 0.4956815242767334, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06285238228738307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03799100033938885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1636957660317421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07186435237526893, + "weighted_rkd_loss": 0.08184788301587105 + }, + { + "epoch": 0.3015075376884422, + "flow_matching_loss": 0.06954732127487659, + "grad_norm": 0.4060089886188507, + "learning_rate": 2.148e-05, + "loss": 0.1457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05899851806461811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034271294809877874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1522683322429657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06954732127487659, + "weighted_rkd_loss": 0.07613416612148285 + }, + { + "epoch": 0.31825795644891125, + "flow_matching_loss": 0.07116162888705731, + "grad_norm": 0.4704577922821045, + "learning_rate": 2.268e-05, + "loss": 0.1464, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05825379975140095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033995861932635305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15050345957279204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07116162888705731, + "weighted_rkd_loss": 0.07525172978639602 + }, + { + "epoch": 0.33500837520938026, + "flow_matching_loss": 0.07152839414775372, + "grad_norm": 0.4596906900405884, + "learning_rate": 2.3880000000000002e-05, + "loss": 0.1431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05577990598976612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03162169083952904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14318150281906128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07152839414775372, + "weighted_rkd_loss": 0.07159075140953064 + }, + { + "epoch": 0.35175879396984927, + "flow_matching_loss": 0.06675980575382709, + "grad_norm": 0.46783247590065, + "learning_rate": 2.508e-05, + "loss": 0.1351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.052671773731708525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0314154077321291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13675895631313323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06675980575382709, + "weighted_rkd_loss": 0.06837947815656661 + }, + { + "epoch": 0.3685092127303183, + "flow_matching_loss": 0.06015884168446064, + "grad_norm": 0.4413231313228607, + "learning_rate": 2.628e-05, + "loss": 0.12, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04683331362903118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02598679680377245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11965342536568642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06015884168446064, + "weighted_rkd_loss": 0.05982671268284321 + }, + { + "epoch": 0.3852596314907873, + "flow_matching_loss": 0.05922078154981136, + "grad_norm": 0.4952040910720825, + "learning_rate": 2.748e-05, + "loss": 0.1187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04629960469901562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02644723802804947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1190464474260807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05922078154981136, + "weighted_rkd_loss": 0.05952322371304035 + }, + { + "epoch": 0.4020100502512563, + "flow_matching_loss": 0.06449385173618793, + "grad_norm": 0.48206594586372375, + "learning_rate": 2.868e-05, + "loss": 0.1224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04490597955882549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025913392566144465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11572535187005997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06449385173618793, + "weighted_rkd_loss": 0.057862675935029986 + }, + { + "epoch": 0.4187604690117253, + "flow_matching_loss": 0.054776575043797494, + "grad_norm": 0.40395382046699524, + "learning_rate": 2.9880000000000002e-05, + "loss": 0.1104, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04352367855608463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024162129126489164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11120948567986488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.054776575043797494, + "weighted_rkd_loss": 0.05560474283993244 + }, + { + "epoch": 0.4355108877721943, + "flow_matching_loss": 0.05850586108863354, + "grad_norm": 0.4056771993637085, + "learning_rate": 2.9999734259635676e-05, + "loss": 0.1118, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041912122443318364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022748018987476824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10657226219773293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05850586108863354, + "weighted_rkd_loss": 0.053286131098866464 + }, + { + "epoch": 0.45226130653266333, + "flow_matching_loss": 0.062202782928943635, + "grad_norm": 0.4282694160938263, + "learning_rate": 2.9998815663057245e-05, + "loss": 0.1095, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036747928336262704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02105743968859315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0945532962679863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.062202782928943635, + "weighted_rkd_loss": 0.04727664813399315 + }, + { + "epoch": 0.46901172529313234, + "flow_matching_loss": 0.051905983313918116, + "grad_norm": 0.39980289340019226, + "learning_rate": 2.999724096969228e-05, + "loss": 0.0978, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03628676719963551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019146752916276454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09172028750181198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.051905983313918116, + "weighted_rkd_loss": 0.04586014375090599 + }, + { + "epoch": 0.48576214405360135, + "flow_matching_loss": 0.05202174261212349, + "grad_norm": 0.42642244696617126, + "learning_rate": 2.9995010248422984e-05, + "loss": 0.0982, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03656212091445923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01933057326823473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0924548164010048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05202174261212349, + "weighted_rkd_loss": 0.0462274082005024 + }, + { + "epoch": 0.5025125628140703, + "flow_matching_loss": 0.048975753411650655, + "grad_norm": 0.3835107982158661, + "learning_rate": 2.999212359682832e-05, + "loss": 0.0927, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034414700232446194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018641455005854367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08747085481882096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.048975753411650655, + "weighted_rkd_loss": 0.04373542740941048 + }, + { + "epoch": 0.5192629815745393, + "flow_matching_loss": 0.05028046742081642, + "grad_norm": 0.3581988513469696, + "learning_rate": 2.9988581141179787e-05, + "loss": 0.0919, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03313282225281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016972948610782624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08323859423398972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05028046742081642, + "weighted_rkd_loss": 0.04161929711699486 + }, + { + "epoch": 0.5360134003350083, + "flow_matching_loss": 0.048362468183040616, + "grad_norm": 0.3743322193622589, + "learning_rate": 2.998438303643587e-05, + "loss": 0.0896, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03231747522950172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017821643501520157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0824565939605236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.048362468183040616, + "weighted_rkd_loss": 0.0412282969802618 + }, + { + "epoch": 0.5527638190954773, + "flow_matching_loss": 0.04962967894971371, + "grad_norm": 0.37111181020736694, + "learning_rate": 2.997952946623527e-05, + "loss": 0.0898, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031945614889264104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016547599248588084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08043882995843887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04962967894971371, + "weighted_rkd_loss": 0.04021941497921944 + }, + { + "epoch": 0.5695142378559463, + "flow_matching_loss": 0.050852949172258376, + "grad_norm": 0.418589323759079, + "learning_rate": 2.9974020642888875e-05, + "loss": 0.0892, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03034352362155914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016099225915968417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07678627483546734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050852949172258376, + "weighted_rkd_loss": 0.03839313741773367 + }, + { + "epoch": 0.5862646566164154, + "flow_matching_loss": 0.05294822007417679, + "grad_norm": 0.4339646100997925, + "learning_rate": 2.9967856807370458e-05, + "loss": 0.0964, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03385459519922733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019144115131348373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08685330525040627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05294822007417679, + "weighted_rkd_loss": 0.043426652625203135 + }, + { + "epoch": 0.6030150753768844, + "flow_matching_loss": 0.04927732534706593, + "grad_norm": 0.39898374676704407, + "learning_rate": 2.996103822930615e-05, + "loss": 0.09, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03226359337568283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01696971897035837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08149690553545952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04927732534706593, + "weighted_rkd_loss": 0.04074845276772976 + }, + { + "epoch": 0.6197654941373534, + "flow_matching_loss": 0.04757112897932529, + "grad_norm": 0.3817186653614044, + "learning_rate": 2.9953565206962653e-05, + "loss": 0.0851, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029814749024808407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015488553512841462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07511805146932601, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04757112897932529, + "weighted_rkd_loss": 0.03755902573466301 + }, + { + "epoch": 0.6365159128978225, + "flow_matching_loss": 0.04780330918729305, + "grad_norm": 0.35171931982040405, + "learning_rate": 2.9945438067234172e-05, + "loss": 0.0839, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028515780717134474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01506363358348608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07209519408643246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04780330918729305, + "weighted_rkd_loss": 0.03604759704321623 + }, + { + "epoch": 0.6532663316582915, + "flow_matching_loss": 0.04430910907685757, + "grad_norm": 0.39219847321510315, + "learning_rate": 2.9936657165628124e-05, + "loss": 0.0793, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027778214029967784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01443298477679491, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06998941153287888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04430910907685757, + "weighted_rkd_loss": 0.03499470576643944 + }, + { + "epoch": 0.6700167504187605, + "flow_matching_loss": 0.04167022332549095, + "grad_norm": 0.3763214945793152, + "learning_rate": 2.992722288624959e-05, + "loss": 0.0763, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02743688691407442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014438783749938011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06931255795061589, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04167022332549095, + "weighted_rkd_loss": 0.034656278975307944 + }, + { + "epoch": 0.6867671691792295, + "flow_matching_loss": 0.04623462222516537, + "grad_norm": 0.3273172974586487, + "learning_rate": 2.9917135641784515e-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": 0.026866826787590982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0142863011918962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06801995486021042, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04623462222516537, + "weighted_rkd_loss": 0.03400997743010521 + }, + { + "epoch": 0.7035175879396985, + "flow_matching_loss": 0.042290421202778816, + "grad_norm": 0.3280941843986511, + "learning_rate": 2.9906395873481655e-05, + "loss": 0.0747, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025736703351140023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013280625827610493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06475403234362602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042290421202778816, + "weighted_rkd_loss": 0.03237701617181301 + }, + { + "epoch": 0.7202680067001676, + "flow_matching_loss": 0.04566530957818031, + "grad_norm": 0.3375310003757477, + "learning_rate": 2.989500405113326e-05, + "loss": 0.0792, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02661526370793581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013754750974476338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06698527745902538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04566530957818031, + "weighted_rkd_loss": 0.03349263872951269 + }, + { + "epoch": 0.7370184254606366, + "flow_matching_loss": 0.041253012418746945, + "grad_norm": 0.36417585611343384, + "learning_rate": 2.9882960673054547e-05, + "loss": 0.0736, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025740651600062846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013157044537365437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06463834792375564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.041253012418746945, + "weighted_rkd_loss": 0.03231917396187782 + }, + { + "epoch": 0.7537688442211056, + "flow_matching_loss": 0.0429521806538105, + "grad_norm": 0.31173717975616455, + "learning_rate": 2.9870266266061886e-05, + "loss": 0.0734, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02408239897340536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012752699572592974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060917497426271436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0429521806538105, + "weighted_rkd_loss": 0.030458748713135718 + }, + { + "epoch": 0.7705192629815746, + "flow_matching_loss": 0.04328401610255241, + "grad_norm": 0.32832983136177063, + "learning_rate": 2.9856921385449766e-05, + "loss": 0.0763, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02588608395308256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01416186960414052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06593403704464436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04328401610255241, + "weighted_rkd_loss": 0.03296701852232218 + }, + { + "epoch": 0.7872696817420436, + "flow_matching_loss": 0.042371243983507154, + "grad_norm": 0.322933167219162, + "learning_rate": 2.9842926614966502e-05, + "loss": 0.0717, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023254065960645675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012198315933346748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0587064478546381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042371243983507154, + "weighted_rkd_loss": 0.02935322392731905 + }, + { + "epoch": 0.8040201005025126, + "flow_matching_loss": 0.036948934011161325, + "grad_norm": 0.37090227007865906, + "learning_rate": 2.982828256678868e-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": 0.023912062868475914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012023937422782183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05984806343913078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.036948934011161325, + "weighted_rkd_loss": 0.02992403171956539 + }, + { + "epoch": 0.8207705192629816, + "flow_matching_loss": 0.04164762869477272, + "grad_norm": 0.2612155079841614, + "learning_rate": 2.9812989881494412e-05, + "loss": 0.0693, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022154004499316215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011067771259695292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05537578016519547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04164762869477272, + "weighted_rkd_loss": 0.027687890082597734 + }, + { + "epoch": 0.8375209380234506, + "flow_matching_loss": 0.041829455830156806, + "grad_norm": 0.3849196434020996, + "learning_rate": 2.9797049228035296e-05, + "loss": 0.0696, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022308938018977643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010920075606554747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05553795211017132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.041829455830156806, + "weighted_rkd_loss": 0.02776897605508566 + }, + { + "epoch": 0.8542713567839196, + "flow_matching_loss": 0.039186670817434786, + "grad_norm": 0.2965879738330841, + "learning_rate": 2.978046130370715e-05, + "loss": 0.0642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02009088359773159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009756325092166663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04993809200823307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.039186670817434786, + "weighted_rkd_loss": 0.024969046004116535 + }, + { + "epoch": 0.8710217755443886, + "flow_matching_loss": 0.03997811581939459, + "grad_norm": 0.3125465512275696, + "learning_rate": 2.9763226834119535e-05, + "loss": 0.0675, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0217224583029747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011559972260147332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05500488877296448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03997811581939459, + "weighted_rkd_loss": 0.02750244438648224 + }, + { + "epoch": 0.8877721943048577, + "flow_matching_loss": 0.037035029381513596, + "grad_norm": 0.28196239471435547, + "learning_rate": 2.9745346573163963e-05, + "loss": 0.0683, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024690970405936243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013126031681895255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06250797249376774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037035029381513596, + "weighted_rkd_loss": 0.03125398624688387 + }, + { + "epoch": 0.9045226130653267, + "flow_matching_loss": 0.03601754531264305, + "grad_norm": 0.3579603135585785, + "learning_rate": 2.9726821302980987e-05, + "loss": 0.0653, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023435172624886036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011736051738262176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05860639661550522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03601754531264305, + "weighted_rkd_loss": 0.02930319830775261 + }, + { + "epoch": 0.9212730318257957, + "flow_matching_loss": 0.035668034479022025, + "grad_norm": 0.33571353554725647, + "learning_rate": 2.9707651833925935e-05, + "loss": 0.0603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01971091693267226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009920926159247756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049342759884893894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035668034479022025, + "weighted_rkd_loss": 0.024671379942446947 + }, + { + "epoch": 0.9380234505862647, + "flow_matching_loss": 0.04034342188388109, + "grad_norm": 0.37223154306411743, + "learning_rate": 2.968783900453348e-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": 0.02227401491254568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010327170602977276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054875200241804124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04034342188388109, + "weighted_rkd_loss": 0.027437600120902062 + }, + { + "epoch": 0.9547738693467337, + "flow_matching_loss": 0.03982900306582451, + "grad_norm": 0.3271760940551758, + "learning_rate": 2.9667383681480968e-05, + "loss": 0.0647, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019855904299765824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010042787250131368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04975459612905979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03982900306582451, + "weighted_rkd_loss": 0.024877298064529894 + }, + { + "epoch": 0.9715242881072027, + "flow_matching_loss": 0.04037307444959879, + "grad_norm": 0.3598499596118927, + "learning_rate": 2.9646286759550493e-05, + "loss": 0.0652, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01992246611043811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009732879279181362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04957781210541725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04037307444959879, + "weighted_rkd_loss": 0.024788906052708625 + }, + { + "epoch": 0.9882747068676717, + "flow_matching_loss": 0.03921057693660259, + "grad_norm": 0.4509306848049164, + "learning_rate": 2.9624549161589764e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018727950938045978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009081155760213733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04653705805540085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03921057693660259, + "weighted_rkd_loss": 0.023268529027700425 + }, + { + "epoch": 1.0050251256281406, + "flow_matching_loss": 0.03464273381978274, + "grad_norm": 0.3406599164009094, + "learning_rate": 2.9602171838471745e-05, + "loss": 0.0614, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02158996183425188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01035784943960607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05353777259588242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03464273381978274, + "weighted_rkd_loss": 0.02676888629794121 + }, + { + "epoch": 1.0217755443886096, + "flow_matching_loss": 0.032238774560391904, + "grad_norm": 0.3566734194755554, + "learning_rate": 2.957915576905303e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019271148182451726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010775866499170662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049318162724375725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032238774560391904, + "weighted_rkd_loss": 0.024659081362187862 + }, + { + "epoch": 1.0385259631490786, + "flow_matching_loss": 0.033852733485400674, + "grad_norm": 0.3310783803462982, + "learning_rate": 2.9555501960131058e-05, + "loss": 0.0578, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019164290837943555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009481828706339002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047810410335659984, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033852733485400674, + "weighted_rkd_loss": 0.023905205167829992 + }, + { + "epoch": 1.0552763819095476, + "flow_matching_loss": 0.03569246102124453, + "grad_norm": 0.29758018255233765, + "learning_rate": 2.9531211446400068e-05, + "loss": 0.0619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020986123383045195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010503490781411528, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052475737407803535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03569246102124453, + "weighted_rkd_loss": 0.026237868703901768 + }, + { + "epoch": 1.0720268006700167, + "flow_matching_loss": 0.03429760187864304, + "grad_norm": 0.34033986926078796, + "learning_rate": 2.9506285290405813e-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": 0.019663255382329225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009310608357191085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048637119308114055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03429760187864304, + "weighted_rkd_loss": 0.024318559654057027 + }, + { + "epoch": 1.0887772194304857, + "flow_matching_loss": 0.03442421518266201, + "grad_norm": 0.3421791195869446, + "learning_rate": 2.948072458249911e-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": 0.018693058285862207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00901120575144887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04639732241630554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03442421518266201, + "weighted_rkd_loss": 0.02319866120815277 + }, + { + "epoch": 1.1055276381909547, + "flow_matching_loss": 0.033443206921219826, + "grad_norm": 0.3156006932258606, + "learning_rate": 2.9454530440788125e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020682687778025864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009997753286734223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05136312879621983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033443206921219826, + "weighted_rkd_loss": 0.025681564398109914 + }, + { + "epoch": 1.1222780569514237, + "flow_matching_loss": 0.035716149397194384, + "grad_norm": 0.3609527349472046, + "learning_rate": 2.9427704011089474e-05, + "loss": 0.0601, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01966800969094038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009427881613373757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04876390062272549, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035716149397194384, + "weighted_rkd_loss": 0.024381950311362744 + }, + { + "epoch": 1.1390284757118927, + "flow_matching_loss": 0.03611791767179966, + "grad_norm": 0.29612571001052856, + "learning_rate": 2.94002464668781e-05, + "loss": 0.0577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017426843661814927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008403770998120308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04325745925307274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03611791767179966, + "weighted_rkd_loss": 0.02162872962653637 + }, + { + "epoch": 1.1557788944723617, + "flow_matching_loss": 0.03446404431015253, + "grad_norm": 0.26498663425445557, + "learning_rate": 2.9372159009235936e-05, + "loss": 0.0559, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01693731788545847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008955044439062476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042829679884016515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03446404431015253, + "weighted_rkd_loss": 0.021414839942008258 + }, + { + "epoch": 1.1725293132328307, + "flow_matching_loss": 0.03775186575949192, + "grad_norm": 0.2350740283727646, + "learning_rate": 2.9343442866799365e-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": 0.01552609372884035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007015585526823997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03806777279824018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03775186575949192, + "weighted_rkd_loss": 0.01903388639912009 + }, + { + "epoch": 1.1892797319932997, + "flow_matching_loss": 0.034802002087235454, + "grad_norm": 0.27234312891960144, + "learning_rate": 2.9314099295705495e-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": 0.01639019064605236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008023013034835458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04080339428037405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034802002087235454, + "weighted_rkd_loss": 0.020401697140187025 + }, + { + "epoch": 1.2060301507537687, + "flow_matching_loss": 0.033095156960189345, + "grad_norm": 0.2403380125761032, + "learning_rate": 2.928412957953718e-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": 0.016718606650829314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00849555665627122, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04193276967853308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033095156960189345, + "weighted_rkd_loss": 0.02096638483926654 + }, + { + "epoch": 1.2227805695142377, + "flow_matching_loss": 0.03338757548481226, + "grad_norm": 0.34521374106407166, + "learning_rate": 2.925353502926689e-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": 0.016283851675689222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007864421885460614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04043212570250034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03338757548481226, + "weighted_rkd_loss": 0.02021606285125017 + }, + { + "epoch": 1.2395309882747068, + "flow_matching_loss": 0.029536503180861474, + "grad_norm": 0.313785195350647, + "learning_rate": 2.922231698319937e-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": 0.015204551909118891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007029514294117689, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03743861857801676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029536503180861474, + "weighted_rkd_loss": 0.01871930928900838 + }, + { + "epoch": 1.2562814070351758, + "flow_matching_loss": 0.031847203895449636, + "grad_norm": 0.31126895546913147, + "learning_rate": 2.9190476806913088e-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": 0.015663769003003834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007163937482982874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038491476140916345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031847203895449636, + "weighted_rkd_loss": 0.019245738070458172 + }, + { + "epoch": 1.2730318257956448, + "flow_matching_loss": 0.03111131377518177, + "grad_norm": 0.24699684977531433, + "learning_rate": 2.9158015893200504e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015276441443711519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007583758374676108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038136640936136244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03111131377518177, + "weighted_rkd_loss": 0.019068320468068122 + }, + { + "epoch": 1.2897822445561138, + "flow_matching_loss": 0.03199950084090233, + "grad_norm": 0.26863768696784973, + "learning_rate": 2.9124935662007137e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014959248062223197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007236627629026771, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03715512380003929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03199950084090233, + "weighted_rkd_loss": 0.018577561900019645 + }, + { + "epoch": 1.3065326633165828, + "flow_matching_loss": 0.031591197848320006, + "grad_norm": 0.27149420976638794, + "learning_rate": 2.9091237560369467e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01730004707351327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007396387634798884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041996481455862524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031591197848320006, + "weighted_rkd_loss": 0.020998240727931262 + }, + { + "epoch": 1.3232830820770518, + "flow_matching_loss": 0.03199263922870159, + "grad_norm": 0.23743563890457153, + "learning_rate": 2.905692306235163e-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": 0.014362255297601222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007487459061667323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03621196895837784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03199263922870159, + "weighted_rkd_loss": 0.01810598447918892 + }, + { + "epoch": 1.3400335008375208, + "flow_matching_loss": 0.032203089259564874, + "grad_norm": 0.26545020937919617, + "learning_rate": 2.9021993668980925e-05, + "loss": 0.0509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015238793008029461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006904208846390247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03738179467618465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032203089259564874, + "weighted_rkd_loss": 0.018690897338092326 + }, + { + "epoch": 1.3567839195979898, + "flow_matching_loss": 0.03435746598988772, + "grad_norm": 0.3084009289741516, + "learning_rate": 2.8986450908182185e-05, + "loss": 0.0519, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014165364857763051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00666372743435204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03499445673078298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03435746598988772, + "weighted_rkd_loss": 0.01749722836539149 + }, + { + "epoch": 1.3735343383584588, + "flow_matching_loss": 0.03177093621343374, + "grad_norm": 0.26156696677207947, + "learning_rate": 2.8950296334710902e-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": 0.01559155946597457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0073688535019755365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03855197243392468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03177093621343374, + "weighted_rkd_loss": 0.01927598621696234 + }, + { + "epoch": 1.3902847571189278, + "flow_matching_loss": 0.03415762279182673, + "grad_norm": 0.3244900107383728, + "learning_rate": 2.8913531530085248e-05, + "loss": 0.054, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015872035082429646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00787778990343213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03962186016142368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03415762279182673, + "weighted_rkd_loss": 0.01981093008071184 + }, + { + "epoch": 1.4070351758793969, + "flow_matching_loss": 0.031543026119470595, + "grad_norm": 0.24670489132404327, + "learning_rate": 2.887615810251687e-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": 0.014911935105919838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007098326459527016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03692219667136669, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031543026119470595, + "weighted_rkd_loss": 0.018461098335683344 + }, + { + "epoch": 1.4237855946398659, + "flow_matching_loss": 0.030299522541463376, + "grad_norm": 0.29750338196754456, + "learning_rate": 2.8838177686840577e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013420206494629383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006395753938704729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033236166834831236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030299522541463376, + "weighted_rkd_loss": 0.016618083417415618 + }, + { + "epoch": 1.4405360134003349, + "flow_matching_loss": 0.03428787123411894, + "grad_norm": 0.2844262421131134, + "learning_rate": 2.8799591944442776e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014772494789212942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007544750696979463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0370897401124239, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03428787123411894, + "weighted_rkd_loss": 0.01854487005621195 + }, + { + "epoch": 1.457286432160804, + "flow_matching_loss": 0.028464125469326973, + "grad_norm": 0.2586044669151306, + "learning_rate": 2.8760402563188848e-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": 0.015234122145920992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007229142403230071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03769738636910915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028464125469326973, + "weighted_rkd_loss": 0.018848693184554576 + }, + { + "epoch": 1.474036850921273, + "flow_matching_loss": 0.03240107372403145, + "grad_norm": 0.3451785743236542, + "learning_rate": 2.8720611257349285e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01441895104944706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007548008835874498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036385910399258134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03240107372403145, + "weighted_rkd_loss": 0.018192955199629067 + }, + { + "epoch": 1.490787269681742, + "flow_matching_loss": 0.031261354126036166, + "grad_norm": 0.2677576243877411, + "learning_rate": 2.868021976752471e-05, + "loss": 0.0504, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015447228122502566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007448036875575781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03834249284118414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031261354126036166, + "weighted_rkd_loss": 0.01917124642059207 + }, + { + "epoch": 1.507537688442211, + "flow_matching_loss": 0.029524659179151057, + "grad_norm": 0.2589884400367737, + "learning_rate": 2.8639229860569738e-05, + "loss": 0.0467, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013584412820637225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007130756136029958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03429958187043667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029524659179151057, + "weighted_rkd_loss": 0.017149790935218335 + }, + { + "epoch": 1.52428810720268, + "flow_matching_loss": 0.030357540026307105, + "grad_norm": 0.2322537899017334, + "learning_rate": 2.8597643329515688e-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": 0.015393463242799044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007288130559027195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03807505648583174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030357540026307105, + "weighted_rkd_loss": 0.01903752824291587 + }, + { + "epoch": 1.541038525963149, + "flow_matching_loss": 0.033645805157721045, + "grad_norm": 0.2303711622953415, + "learning_rate": 2.855546199349214e-05, + "loss": 0.0525, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0153488309122622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070605908520519735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037758252769708636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033645805157721045, + "weighted_rkd_loss": 0.018879126384854318 + }, + { + "epoch": 1.557788944723618, + "flow_matching_loss": 0.029417644068598748, + "grad_norm": 0.33168652653694153, + "learning_rate": 2.85126876976474e-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": 0.013482074066996575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007291349256411195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03425549734383822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029417644068598748, + "weighted_rkd_loss": 0.01712774867191911 + }, + { + "epoch": 1.574539363484087, + "flow_matching_loss": 0.02936752066016197, + "grad_norm": 0.32166430354118347, + "learning_rate": 2.8469322313067726e-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": 0.014363228995352983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006481660483404994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035208118706941606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02936752066016197, + "weighted_rkd_loss": 0.017604059353470803 + }, + { + "epoch": 1.591289782244556, + "flow_matching_loss": 0.030744343996047974, + "grad_norm": 0.2652220129966736, + "learning_rate": 2.8425367736695522e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012823993526399135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006567410938441753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03221539761871099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030744343996047974, + "weighted_rkd_loss": 0.016107698809355496 + }, + { + "epoch": 1.608040201005025, + "flow_matching_loss": 0.027458759397268294, + "grad_norm": 0.2823118269443512, + "learning_rate": 2.838082589124636e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014246038906276227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006480111461132765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03497218918055296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027458759397268294, + "weighted_rkd_loss": 0.01748609459027648 + }, + { + "epoch": 1.624790619765494, + "flow_matching_loss": 0.02682869080454111, + "grad_norm": 0.27871882915496826, + "learning_rate": 2.8335698725124843e-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": 0.015229499991983175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0072166129481047395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037675613351166246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02682869080454111, + "weighted_rkd_loss": 0.018837806675583123 + }, + { + "epoch": 1.641541038525963, + "flow_matching_loss": 0.027610163576900958, + "grad_norm": 0.27080997824668884, + "learning_rate": 2.8289988212339422e-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": 0.014654815848916768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007433690736070276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03674332220107317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027610163576900958, + "weighted_rkd_loss": 0.018371661100536583 + }, + { + "epoch": 1.658291457286432, + "flow_matching_loss": 0.026477544382214547, + "grad_norm": 0.29053348302841187, + "learning_rate": 2.8243696352415997e-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": 0.014315103739500045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006944430945441127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035574638098478314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026477544382214547, + "weighted_rkd_loss": 0.017787319049239157 + }, + { + "epoch": 1.675041876046901, + "flow_matching_loss": 0.03149757422506809, + "grad_norm": 0.24316534399986267, + "learning_rate": 2.8196825170310495e-05, + "loss": 0.0489, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013985770475119352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006879748497158289, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034851289354264736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03149757422506809, + "weighted_rkd_loss": 0.017425644677132368 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..27700cac5d292d4d42897b3c2e80bdc58e55c60e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5719731235192254, + -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-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a3758b2dac9cab1274647a0a3e4647bb197ec7fc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6f702e66e76c9c2a5773af5ec6b03a2aabb911ab03f04ce0b6ef1d2a8dca66a +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..e5e10bca0e32678a4d2bc39bbddacd4bdb090c5e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-3000/trainer_state.json @@ -0,0 +1,13834 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 5.025125628140704, + "eval_steps": 500, + "global_step": 3000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.01675041876046901, + "flow_matching_loss": 0.20356450378894805, + "grad_norm": 1.0490323305130005, + "learning_rate": 1.08e-06, + "loss": 0.4004, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1398262470960617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11395216882228851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.39360466301441194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.20356450378894805, + "weighted_rkd_loss": 0.19680233150720597 + }, + { + "epoch": 0.03350083752093802, + "flow_matching_loss": 0.21135056614875794, + "grad_norm": 0.8279632925987244, + "learning_rate": 2.28e-06, + "loss": 0.3916, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12995845153927804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10064354985952377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3605604529380798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.21135056614875794, + "weighted_rkd_loss": 0.1802802264690399 + }, + { + "epoch": 0.05025125628140704, + "flow_matching_loss": 0.19959198087453842, + "grad_norm": 0.8481504321098328, + "learning_rate": 3.48e-06, + "loss": 0.3868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13302332013845444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1083116926252842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.374358332157135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.19959198087453842, + "weighted_rkd_loss": 0.1871791660785675 + }, + { + "epoch": 0.06700167504187604, + "flow_matching_loss": 0.21401427686214447, + "grad_norm": 1.1905933618545532, + "learning_rate": 4.68e-06, + "loss": 0.3892, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12562137991189956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09915168732404708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3503944456577301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.21401427686214447, + "weighted_rkd_loss": 0.17519722282886505 + }, + { + "epoch": 0.08375209380234507, + "flow_matching_loss": 0.18060285598039627, + "grad_norm": 0.9187169075012207, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11310266405344009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08473661690950393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3109419494867325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.18060285598039627, + "weighted_rkd_loss": 0.15547097474336624 + }, + { + "epoch": 0.10050251256281408, + "flow_matching_loss": 0.1746109426021576, + "grad_norm": 0.9370708465576172, + "learning_rate": 7.08e-06, + "loss": 0.3298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11362355500459671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08309690579771996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.31034401059150696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1746109426021576, + "weighted_rkd_loss": 0.15517200529575348 + }, + { + "epoch": 0.11725293132328309, + "flow_matching_loss": 0.16165669709444047, + "grad_norm": 0.7368708848953247, + "learning_rate": 8.28e-06, + "loss": 0.3148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11314839944243431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07991818860173225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30621498823165894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.16165669709444047, + "weighted_rkd_loss": 0.15310749411582947 + }, + { + "epoch": 0.13400335008375208, + "flow_matching_loss": 0.15373115316033364, + "grad_norm": 0.8529441356658936, + "learning_rate": 9.48e-06, + "loss": 0.2947, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10446386188268661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0729650903493166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.28189281225204466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.15373115316033364, + "weighted_rkd_loss": 0.14094640612602233 + }, + { + "epoch": 0.1507537688442211, + "flow_matching_loss": 0.12661252617836, + "grad_norm": 0.7936683893203735, + "learning_rate": 1.068e-05, + "loss": 0.2582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09823484048247337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06663991399109363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2631095930933952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.12661252617836, + "weighted_rkd_loss": 0.1315547965466976 + }, + { + "epoch": 0.16750418760469013, + "flow_matching_loss": 0.10844202637672425, + "grad_norm": 0.6985727548599243, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.2325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09290647432208061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.062390904128551486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24820385426282882, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.10844202637672425, + "weighted_rkd_loss": 0.12410192713141441 + }, + { + "epoch": 0.18425460636515914, + "flow_matching_loss": 0.09378867372870445, + "grad_norm": 0.6144883036613464, + "learning_rate": 1.308e-05, + "loss": 0.2126, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08944455981254577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05871000774204731, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23759912848472595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09378867372870445, + "weighted_rkd_loss": 0.11879956424236297 + }, + { + "epoch": 0.20100502512562815, + "flow_matching_loss": 0.08636006340384483, + "grad_norm": 0.519714891910553, + "learning_rate": 1.428e-05, + "loss": 0.1946, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08254799991846085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0513349212706089, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21643092334270478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08636006340384483, + "weighted_rkd_loss": 0.10821546167135239 + }, + { + "epoch": 0.21775544388609716, + "flow_matching_loss": 0.07800828218460083, + "grad_norm": 0.6437940001487732, + "learning_rate": 1.548e-05, + "loss": 0.1787, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07631000094115734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.048831694200634955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20145169496536255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07800828218460083, + "weighted_rkd_loss": 0.10072584748268128 + }, + { + "epoch": 0.23450586264656617, + "flow_matching_loss": 0.08341701701283455, + "grad_norm": 0.6835018992424011, + "learning_rate": 1.6680000000000003e-05, + "loss": 0.1816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0760519951581955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04422654993832111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19633053839206696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08341701701283455, + "weighted_rkd_loss": 0.09816526919603348 + }, + { + "epoch": 0.25125628140703515, + "flow_matching_loss": 0.07501983791589736, + "grad_norm": 0.6060196161270142, + "learning_rate": 1.7879999999999998e-05, + "loss": 0.1644, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06893355883657933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04092012867331505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17878724485635758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07501983791589736, + "weighted_rkd_loss": 0.08939362242817879 + }, + { + "epoch": 0.26800670016750416, + "flow_matching_loss": 0.07575965747237205, + "grad_norm": 0.5369148254394531, + "learning_rate": 1.908e-05, + "loss": 0.1611, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06494721546769142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.040794138610363004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17068857252597808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07575965747237205, + "weighted_rkd_loss": 0.08534428626298904 + }, + { + "epoch": 0.2847571189279732, + "flow_matching_loss": 0.07186435237526893, + "grad_norm": 0.4956815242767334, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06285238228738307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03799100033938885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1636957660317421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07186435237526893, + "weighted_rkd_loss": 0.08184788301587105 + }, + { + "epoch": 0.3015075376884422, + "flow_matching_loss": 0.06954732127487659, + "grad_norm": 0.4060089886188507, + "learning_rate": 2.148e-05, + "loss": 0.1457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05899851806461811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034271294809877874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1522683322429657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06954732127487659, + "weighted_rkd_loss": 0.07613416612148285 + }, + { + "epoch": 0.31825795644891125, + "flow_matching_loss": 0.07116162888705731, + "grad_norm": 0.4704577922821045, + "learning_rate": 2.268e-05, + "loss": 0.1464, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05825379975140095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033995861932635305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15050345957279204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07116162888705731, + "weighted_rkd_loss": 0.07525172978639602 + }, + { + "epoch": 0.33500837520938026, + "flow_matching_loss": 0.07152839414775372, + "grad_norm": 0.4596906900405884, + "learning_rate": 2.3880000000000002e-05, + "loss": 0.1431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05577990598976612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03162169083952904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14318150281906128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07152839414775372, + "weighted_rkd_loss": 0.07159075140953064 + }, + { + "epoch": 0.35175879396984927, + "flow_matching_loss": 0.06675980575382709, + "grad_norm": 0.46783247590065, + "learning_rate": 2.508e-05, + "loss": 0.1351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.052671773731708525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0314154077321291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13675895631313323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06675980575382709, + "weighted_rkd_loss": 0.06837947815656661 + }, + { + "epoch": 0.3685092127303183, + "flow_matching_loss": 0.06015884168446064, + "grad_norm": 0.4413231313228607, + "learning_rate": 2.628e-05, + "loss": 0.12, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04683331362903118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02598679680377245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11965342536568642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06015884168446064, + "weighted_rkd_loss": 0.05982671268284321 + }, + { + "epoch": 0.3852596314907873, + "flow_matching_loss": 0.05922078154981136, + "grad_norm": 0.4952040910720825, + "learning_rate": 2.748e-05, + "loss": 0.1187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04629960469901562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02644723802804947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1190464474260807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05922078154981136, + "weighted_rkd_loss": 0.05952322371304035 + }, + { + "epoch": 0.4020100502512563, + "flow_matching_loss": 0.06449385173618793, + "grad_norm": 0.48206594586372375, + "learning_rate": 2.868e-05, + "loss": 0.1224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04490597955882549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025913392566144465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11572535187005997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06449385173618793, + "weighted_rkd_loss": 0.057862675935029986 + }, + { + "epoch": 0.4187604690117253, + "flow_matching_loss": 0.054776575043797494, + "grad_norm": 0.40395382046699524, + "learning_rate": 2.9880000000000002e-05, + "loss": 0.1104, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04352367855608463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024162129126489164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11120948567986488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.054776575043797494, + "weighted_rkd_loss": 0.05560474283993244 + }, + { + "epoch": 0.4355108877721943, + "flow_matching_loss": 0.05850586108863354, + "grad_norm": 0.4056771993637085, + "learning_rate": 2.9999734259635676e-05, + "loss": 0.1118, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041912122443318364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022748018987476824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10657226219773293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05850586108863354, + "weighted_rkd_loss": 0.053286131098866464 + }, + { + "epoch": 0.45226130653266333, + "flow_matching_loss": 0.062202782928943635, + "grad_norm": 0.4282694160938263, + "learning_rate": 2.9998815663057245e-05, + "loss": 0.1095, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036747928336262704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02105743968859315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0945532962679863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.062202782928943635, + "weighted_rkd_loss": 0.04727664813399315 + }, + { + "epoch": 0.46901172529313234, + "flow_matching_loss": 0.051905983313918116, + "grad_norm": 0.39980289340019226, + "learning_rate": 2.999724096969228e-05, + "loss": 0.0978, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03628676719963551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019146752916276454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09172028750181198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.051905983313918116, + "weighted_rkd_loss": 0.04586014375090599 + }, + { + "epoch": 0.48576214405360135, + "flow_matching_loss": 0.05202174261212349, + "grad_norm": 0.42642244696617126, + "learning_rate": 2.9995010248422984e-05, + "loss": 0.0982, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03656212091445923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01933057326823473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0924548164010048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05202174261212349, + "weighted_rkd_loss": 0.0462274082005024 + }, + { + "epoch": 0.5025125628140703, + "flow_matching_loss": 0.048975753411650655, + "grad_norm": 0.3835107982158661, + "learning_rate": 2.999212359682832e-05, + "loss": 0.0927, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034414700232446194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018641455005854367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08747085481882096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.048975753411650655, + "weighted_rkd_loss": 0.04373542740941048 + }, + { + "epoch": 0.5192629815745393, + "flow_matching_loss": 0.05028046742081642, + "grad_norm": 0.3581988513469696, + "learning_rate": 2.9988581141179787e-05, + "loss": 0.0919, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03313282225281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016972948610782624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08323859423398972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05028046742081642, + "weighted_rkd_loss": 0.04161929711699486 + }, + { + "epoch": 0.5360134003350083, + "flow_matching_loss": 0.048362468183040616, + "grad_norm": 0.3743322193622589, + "learning_rate": 2.998438303643587e-05, + "loss": 0.0896, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03231747522950172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017821643501520157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0824565939605236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.048362468183040616, + "weighted_rkd_loss": 0.0412282969802618 + }, + { + "epoch": 0.5527638190954773, + "flow_matching_loss": 0.04962967894971371, + "grad_norm": 0.37111181020736694, + "learning_rate": 2.997952946623527e-05, + "loss": 0.0898, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031945614889264104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016547599248588084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08043882995843887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04962967894971371, + "weighted_rkd_loss": 0.04021941497921944 + }, + { + "epoch": 0.5695142378559463, + "flow_matching_loss": 0.050852949172258376, + "grad_norm": 0.418589323759079, + "learning_rate": 2.9974020642888875e-05, + "loss": 0.0892, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03034352362155914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016099225915968417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07678627483546734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050852949172258376, + "weighted_rkd_loss": 0.03839313741773367 + }, + { + "epoch": 0.5862646566164154, + "flow_matching_loss": 0.05294822007417679, + "grad_norm": 0.4339646100997925, + "learning_rate": 2.9967856807370458e-05, + "loss": 0.0964, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03385459519922733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019144115131348373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08685330525040627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05294822007417679, + "weighted_rkd_loss": 0.043426652625203135 + }, + { + "epoch": 0.6030150753768844, + "flow_matching_loss": 0.04927732534706593, + "grad_norm": 0.39898374676704407, + "learning_rate": 2.996103822930615e-05, + "loss": 0.09, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03226359337568283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01696971897035837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08149690553545952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04927732534706593, + "weighted_rkd_loss": 0.04074845276772976 + }, + { + "epoch": 0.6197654941373534, + "flow_matching_loss": 0.04757112897932529, + "grad_norm": 0.3817186653614044, + "learning_rate": 2.9953565206962653e-05, + "loss": 0.0851, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029814749024808407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015488553512841462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07511805146932601, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04757112897932529, + "weighted_rkd_loss": 0.03755902573466301 + }, + { + "epoch": 0.6365159128978225, + "flow_matching_loss": 0.04780330918729305, + "grad_norm": 0.35171931982040405, + "learning_rate": 2.9945438067234172e-05, + "loss": 0.0839, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028515780717134474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01506363358348608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07209519408643246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04780330918729305, + "weighted_rkd_loss": 0.03604759704321623 + }, + { + "epoch": 0.6532663316582915, + "flow_matching_loss": 0.04430910907685757, + "grad_norm": 0.39219847321510315, + "learning_rate": 2.9936657165628124e-05, + "loss": 0.0793, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027778214029967784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01443298477679491, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06998941153287888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04430910907685757, + "weighted_rkd_loss": 0.03499470576643944 + }, + { + "epoch": 0.6700167504187605, + "flow_matching_loss": 0.04167022332549095, + "grad_norm": 0.3763214945793152, + "learning_rate": 2.992722288624959e-05, + "loss": 0.0763, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02743688691407442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014438783749938011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06931255795061589, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04167022332549095, + "weighted_rkd_loss": 0.034656278975307944 + }, + { + "epoch": 0.6867671691792295, + "flow_matching_loss": 0.04623462222516537, + "grad_norm": 0.3273172974586487, + "learning_rate": 2.9917135641784515e-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": 0.026866826787590982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0142863011918962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06801995486021042, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04623462222516537, + "weighted_rkd_loss": 0.03400997743010521 + }, + { + "epoch": 0.7035175879396985, + "flow_matching_loss": 0.042290421202778816, + "grad_norm": 0.3280941843986511, + "learning_rate": 2.9906395873481655e-05, + "loss": 0.0747, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025736703351140023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013280625827610493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06475403234362602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042290421202778816, + "weighted_rkd_loss": 0.03237701617181301 + }, + { + "epoch": 0.7202680067001676, + "flow_matching_loss": 0.04566530957818031, + "grad_norm": 0.3375310003757477, + "learning_rate": 2.989500405113326e-05, + "loss": 0.0792, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02661526370793581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013754750974476338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06698527745902538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04566530957818031, + "weighted_rkd_loss": 0.03349263872951269 + }, + { + "epoch": 0.7370184254606366, + "flow_matching_loss": 0.041253012418746945, + "grad_norm": 0.36417585611343384, + "learning_rate": 2.9882960673054547e-05, + "loss": 0.0736, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025740651600062846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013157044537365437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06463834792375564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.041253012418746945, + "weighted_rkd_loss": 0.03231917396187782 + }, + { + "epoch": 0.7537688442211056, + "flow_matching_loss": 0.0429521806538105, + "grad_norm": 0.31173717975616455, + "learning_rate": 2.9870266266061886e-05, + "loss": 0.0734, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02408239897340536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012752699572592974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060917497426271436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0429521806538105, + "weighted_rkd_loss": 0.030458748713135718 + }, + { + "epoch": 0.7705192629815746, + "flow_matching_loss": 0.04328401610255241, + "grad_norm": 0.32832983136177063, + "learning_rate": 2.9856921385449766e-05, + "loss": 0.0763, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02588608395308256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01416186960414052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06593403704464436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04328401610255241, + "weighted_rkd_loss": 0.03296701852232218 + }, + { + "epoch": 0.7872696817420436, + "flow_matching_loss": 0.042371243983507154, + "grad_norm": 0.322933167219162, + "learning_rate": 2.9842926614966502e-05, + "loss": 0.0717, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023254065960645675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012198315933346748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0587064478546381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042371243983507154, + "weighted_rkd_loss": 0.02935322392731905 + }, + { + "epoch": 0.8040201005025126, + "flow_matching_loss": 0.036948934011161325, + "grad_norm": 0.37090227007865906, + "learning_rate": 2.982828256678868e-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": 0.023912062868475914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012023937422782183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05984806343913078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.036948934011161325, + "weighted_rkd_loss": 0.02992403171956539 + }, + { + "epoch": 0.8207705192629816, + "flow_matching_loss": 0.04164762869477272, + "grad_norm": 0.2612155079841614, + "learning_rate": 2.9812989881494412e-05, + "loss": 0.0693, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022154004499316215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011067771259695292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05537578016519547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04164762869477272, + "weighted_rkd_loss": 0.027687890082597734 + }, + { + "epoch": 0.8375209380234506, + "flow_matching_loss": 0.041829455830156806, + "grad_norm": 0.3849196434020996, + "learning_rate": 2.9797049228035296e-05, + "loss": 0.0696, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022308938018977643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010920075606554747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05553795211017132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.041829455830156806, + "weighted_rkd_loss": 0.02776897605508566 + }, + { + "epoch": 0.8542713567839196, + "flow_matching_loss": 0.039186670817434786, + "grad_norm": 0.2965879738330841, + "learning_rate": 2.978046130370715e-05, + "loss": 0.0642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02009088359773159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009756325092166663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04993809200823307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.039186670817434786, + "weighted_rkd_loss": 0.024969046004116535 + }, + { + "epoch": 0.8710217755443886, + "flow_matching_loss": 0.03997811581939459, + "grad_norm": 0.3125465512275696, + "learning_rate": 2.9763226834119535e-05, + "loss": 0.0675, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0217224583029747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011559972260147332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05500488877296448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03997811581939459, + "weighted_rkd_loss": 0.02750244438648224 + }, + { + "epoch": 0.8877721943048577, + "flow_matching_loss": 0.037035029381513596, + "grad_norm": 0.28196239471435547, + "learning_rate": 2.9745346573163963e-05, + "loss": 0.0683, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024690970405936243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013126031681895255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06250797249376774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037035029381513596, + "weighted_rkd_loss": 0.03125398624688387 + }, + { + "epoch": 0.9045226130653267, + "flow_matching_loss": 0.03601754531264305, + "grad_norm": 0.3579603135585785, + "learning_rate": 2.9726821302980987e-05, + "loss": 0.0653, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023435172624886036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011736051738262176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05860639661550522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03601754531264305, + "weighted_rkd_loss": 0.02930319830775261 + }, + { + "epoch": 0.9212730318257957, + "flow_matching_loss": 0.035668034479022025, + "grad_norm": 0.33571353554725647, + "learning_rate": 2.9707651833925935e-05, + "loss": 0.0603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01971091693267226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009920926159247756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049342759884893894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035668034479022025, + "weighted_rkd_loss": 0.024671379942446947 + }, + { + "epoch": 0.9380234505862647, + "flow_matching_loss": 0.04034342188388109, + "grad_norm": 0.37223154306411743, + "learning_rate": 2.968783900453348e-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": 0.02227401491254568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010327170602977276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054875200241804124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04034342188388109, + "weighted_rkd_loss": 0.027437600120902062 + }, + { + "epoch": 0.9547738693467337, + "flow_matching_loss": 0.03982900306582451, + "grad_norm": 0.3271760940551758, + "learning_rate": 2.9667383681480968e-05, + "loss": 0.0647, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019855904299765824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010042787250131368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04975459612905979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03982900306582451, + "weighted_rkd_loss": 0.024877298064529894 + }, + { + "epoch": 0.9715242881072027, + "flow_matching_loss": 0.04037307444959879, + "grad_norm": 0.3598499596118927, + "learning_rate": 2.9646286759550493e-05, + "loss": 0.0652, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01992246611043811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009732879279181362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04957781210541725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04037307444959879, + "weighted_rkd_loss": 0.024788906052708625 + }, + { + "epoch": 0.9882747068676717, + "flow_matching_loss": 0.03921057693660259, + "grad_norm": 0.4509306848049164, + "learning_rate": 2.9624549161589764e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018727950938045978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009081155760213733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04653705805540085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03921057693660259, + "weighted_rkd_loss": 0.023268529027700425 + }, + { + "epoch": 1.0050251256281406, + "flow_matching_loss": 0.03464273381978274, + "grad_norm": 0.3406599164009094, + "learning_rate": 2.9602171838471745e-05, + "loss": 0.0614, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02158996183425188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01035784943960607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05353777259588242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03464273381978274, + "weighted_rkd_loss": 0.02676888629794121 + }, + { + "epoch": 1.0217755443886096, + "flow_matching_loss": 0.032238774560391904, + "grad_norm": 0.3566734194755554, + "learning_rate": 2.957915576905303e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019271148182451726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010775866499170662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049318162724375725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032238774560391904, + "weighted_rkd_loss": 0.024659081362187862 + }, + { + "epoch": 1.0385259631490786, + "flow_matching_loss": 0.033852733485400674, + "grad_norm": 0.3310783803462982, + "learning_rate": 2.9555501960131058e-05, + "loss": 0.0578, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019164290837943555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009481828706339002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047810410335659984, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033852733485400674, + "weighted_rkd_loss": 0.023905205167829992 + }, + { + "epoch": 1.0552763819095476, + "flow_matching_loss": 0.03569246102124453, + "grad_norm": 0.29758018255233765, + "learning_rate": 2.9531211446400068e-05, + "loss": 0.0619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020986123383045195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010503490781411528, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052475737407803535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03569246102124453, + "weighted_rkd_loss": 0.026237868703901768 + }, + { + "epoch": 1.0720268006700167, + "flow_matching_loss": 0.03429760187864304, + "grad_norm": 0.34033986926078796, + "learning_rate": 2.9506285290405813e-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": 0.019663255382329225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009310608357191085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048637119308114055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03429760187864304, + "weighted_rkd_loss": 0.024318559654057027 + }, + { + "epoch": 1.0887772194304857, + "flow_matching_loss": 0.03442421518266201, + "grad_norm": 0.3421791195869446, + "learning_rate": 2.948072458249911e-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": 0.018693058285862207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00901120575144887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04639732241630554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03442421518266201, + "weighted_rkd_loss": 0.02319866120815277 + }, + { + "epoch": 1.1055276381909547, + "flow_matching_loss": 0.033443206921219826, + "grad_norm": 0.3156006932258606, + "learning_rate": 2.9454530440788125e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020682687778025864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009997753286734223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05136312879621983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033443206921219826, + "weighted_rkd_loss": 0.025681564398109914 + }, + { + "epoch": 1.1222780569514237, + "flow_matching_loss": 0.035716149397194384, + "grad_norm": 0.3609527349472046, + "learning_rate": 2.9427704011089474e-05, + "loss": 0.0601, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01966800969094038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009427881613373757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04876390062272549, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035716149397194384, + "weighted_rkd_loss": 0.024381950311362744 + }, + { + "epoch": 1.1390284757118927, + "flow_matching_loss": 0.03611791767179966, + "grad_norm": 0.29612571001052856, + "learning_rate": 2.94002464668781e-05, + "loss": 0.0577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017426843661814927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008403770998120308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04325745925307274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03611791767179966, + "weighted_rkd_loss": 0.02162872962653637 + }, + { + "epoch": 1.1557788944723617, + "flow_matching_loss": 0.03446404431015253, + "grad_norm": 0.26498663425445557, + "learning_rate": 2.9372159009235936e-05, + "loss": 0.0559, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01693731788545847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008955044439062476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042829679884016515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03446404431015253, + "weighted_rkd_loss": 0.021414839942008258 + }, + { + "epoch": 1.1725293132328307, + "flow_matching_loss": 0.03775186575949192, + "grad_norm": 0.2350740283727646, + "learning_rate": 2.9343442866799365e-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": 0.01552609372884035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007015585526823997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03806777279824018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03775186575949192, + "weighted_rkd_loss": 0.01903388639912009 + }, + { + "epoch": 1.1892797319932997, + "flow_matching_loss": 0.034802002087235454, + "grad_norm": 0.27234312891960144, + "learning_rate": 2.9314099295705495e-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": 0.01639019064605236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008023013034835458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04080339428037405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034802002087235454, + "weighted_rkd_loss": 0.020401697140187025 + }, + { + "epoch": 1.2060301507537687, + "flow_matching_loss": 0.033095156960189345, + "grad_norm": 0.2403380125761032, + "learning_rate": 2.928412957953718e-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": 0.016718606650829314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00849555665627122, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04193276967853308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033095156960189345, + "weighted_rkd_loss": 0.02096638483926654 + }, + { + "epoch": 1.2227805695142377, + "flow_matching_loss": 0.03338757548481226, + "grad_norm": 0.34521374106407166, + "learning_rate": 2.925353502926689e-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": 0.016283851675689222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007864421885460614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04043212570250034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03338757548481226, + "weighted_rkd_loss": 0.02021606285125017 + }, + { + "epoch": 1.2395309882747068, + "flow_matching_loss": 0.029536503180861474, + "grad_norm": 0.313785195350647, + "learning_rate": 2.922231698319937e-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": 0.015204551909118891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007029514294117689, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03743861857801676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029536503180861474, + "weighted_rkd_loss": 0.01871930928900838 + }, + { + "epoch": 1.2562814070351758, + "flow_matching_loss": 0.031847203895449636, + "grad_norm": 0.31126895546913147, + "learning_rate": 2.9190476806913088e-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": 0.015663769003003834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007163937482982874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038491476140916345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031847203895449636, + "weighted_rkd_loss": 0.019245738070458172 + }, + { + "epoch": 1.2730318257956448, + "flow_matching_loss": 0.03111131377518177, + "grad_norm": 0.24699684977531433, + "learning_rate": 2.9158015893200504e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015276441443711519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007583758374676108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038136640936136244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03111131377518177, + "weighted_rkd_loss": 0.019068320468068122 + }, + { + "epoch": 1.2897822445561138, + "flow_matching_loss": 0.03199950084090233, + "grad_norm": 0.26863768696784973, + "learning_rate": 2.9124935662007137e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014959248062223197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007236627629026771, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03715512380003929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03199950084090233, + "weighted_rkd_loss": 0.018577561900019645 + }, + { + "epoch": 1.3065326633165828, + "flow_matching_loss": 0.031591197848320006, + "grad_norm": 0.27149420976638794, + "learning_rate": 2.9091237560369467e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01730004707351327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007396387634798884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041996481455862524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031591197848320006, + "weighted_rkd_loss": 0.020998240727931262 + }, + { + "epoch": 1.3232830820770518, + "flow_matching_loss": 0.03199263922870159, + "grad_norm": 0.23743563890457153, + "learning_rate": 2.905692306235163e-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": 0.014362255297601222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007487459061667323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03621196895837784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03199263922870159, + "weighted_rkd_loss": 0.01810598447918892 + }, + { + "epoch": 1.3400335008375208, + "flow_matching_loss": 0.032203089259564874, + "grad_norm": 0.26545020937919617, + "learning_rate": 2.9021993668980925e-05, + "loss": 0.0509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015238793008029461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006904208846390247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03738179467618465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032203089259564874, + "weighted_rkd_loss": 0.018690897338092326 + }, + { + "epoch": 1.3567839195979898, + "flow_matching_loss": 0.03435746598988772, + "grad_norm": 0.3084009289741516, + "learning_rate": 2.8986450908182185e-05, + "loss": 0.0519, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014165364857763051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00666372743435204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03499445673078298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03435746598988772, + "weighted_rkd_loss": 0.01749722836539149 + }, + { + "epoch": 1.3735343383584588, + "flow_matching_loss": 0.03177093621343374, + "grad_norm": 0.26156696677207947, + "learning_rate": 2.8950296334710902e-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": 0.01559155946597457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0073688535019755365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03855197243392468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03177093621343374, + "weighted_rkd_loss": 0.01927598621696234 + }, + { + "epoch": 1.3902847571189278, + "flow_matching_loss": 0.03415762279182673, + "grad_norm": 0.3244900107383728, + "learning_rate": 2.8913531530085248e-05, + "loss": 0.054, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015872035082429646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00787778990343213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03962186016142368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03415762279182673, + "weighted_rkd_loss": 0.01981093008071184 + }, + { + "epoch": 1.4070351758793969, + "flow_matching_loss": 0.031543026119470595, + "grad_norm": 0.24670489132404327, + "learning_rate": 2.887615810251687e-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": 0.014911935105919838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007098326459527016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03692219667136669, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031543026119470595, + "weighted_rkd_loss": 0.018461098335683344 + }, + { + "epoch": 1.4237855946398659, + "flow_matching_loss": 0.030299522541463376, + "grad_norm": 0.29750338196754456, + "learning_rate": 2.8838177686840577e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013420206494629383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006395753938704729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033236166834831236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030299522541463376, + "weighted_rkd_loss": 0.016618083417415618 + }, + { + "epoch": 1.4405360134003349, + "flow_matching_loss": 0.03428787123411894, + "grad_norm": 0.2844262421131134, + "learning_rate": 2.8799591944442776e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014772494789212942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007544750696979463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0370897401124239, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03428787123411894, + "weighted_rkd_loss": 0.01854487005621195 + }, + { + "epoch": 1.457286432160804, + "flow_matching_loss": 0.028464125469326973, + "grad_norm": 0.2586044669151306, + "learning_rate": 2.8760402563188848e-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": 0.015234122145920992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007229142403230071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03769738636910915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028464125469326973, + "weighted_rkd_loss": 0.018848693184554576 + }, + { + "epoch": 1.474036850921273, + "flow_matching_loss": 0.03240107372403145, + "grad_norm": 0.3451785743236542, + "learning_rate": 2.8720611257349285e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01441895104944706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007548008835874498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036385910399258134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03240107372403145, + "weighted_rkd_loss": 0.018192955199629067 + }, + { + "epoch": 1.490787269681742, + "flow_matching_loss": 0.031261354126036166, + "grad_norm": 0.2677576243877411, + "learning_rate": 2.868021976752471e-05, + "loss": 0.0504, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015447228122502566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007448036875575781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03834249284118414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031261354126036166, + "weighted_rkd_loss": 0.01917124642059207 + }, + { + "epoch": 1.507537688442211, + "flow_matching_loss": 0.029524659179151057, + "grad_norm": 0.2589884400367737, + "learning_rate": 2.8639229860569738e-05, + "loss": 0.0467, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013584412820637225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007130756136029958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03429958187043667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029524659179151057, + "weighted_rkd_loss": 0.017149790935218335 + }, + { + "epoch": 1.52428810720268, + "flow_matching_loss": 0.030357540026307105, + "grad_norm": 0.2322537899017334, + "learning_rate": 2.8597643329515688e-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": 0.015393463242799044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007288130559027195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03807505648583174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030357540026307105, + "weighted_rkd_loss": 0.01903752824291587 + }, + { + "epoch": 1.541038525963149, + "flow_matching_loss": 0.033645805157721045, + "grad_norm": 0.2303711622953415, + "learning_rate": 2.855546199349214e-05, + "loss": 0.0525, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0153488309122622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070605908520519735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037758252769708636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033645805157721045, + "weighted_rkd_loss": 0.018879126384854318 + }, + { + "epoch": 1.557788944723618, + "flow_matching_loss": 0.029417644068598748, + "grad_norm": 0.33168652653694153, + "learning_rate": 2.85126876976474e-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": 0.013482074066996575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007291349256411195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03425549734383822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029417644068598748, + "weighted_rkd_loss": 0.01712774867191911 + }, + { + "epoch": 1.574539363484087, + "flow_matching_loss": 0.02936752066016197, + "grad_norm": 0.32166430354118347, + "learning_rate": 2.8469322313067726e-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": 0.014363228995352983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006481660483404994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035208118706941606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02936752066016197, + "weighted_rkd_loss": 0.017604059353470803 + }, + { + "epoch": 1.591289782244556, + "flow_matching_loss": 0.030744343996047974, + "grad_norm": 0.2652220129966736, + "learning_rate": 2.8425367736695522e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012823993526399135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006567410938441753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03221539761871099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030744343996047974, + "weighted_rkd_loss": 0.016107698809355496 + }, + { + "epoch": 1.608040201005025, + "flow_matching_loss": 0.027458759397268294, + "grad_norm": 0.2823118269443512, + "learning_rate": 2.838082589124636e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014246038906276227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006480111461132765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03497218918055296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027458759397268294, + "weighted_rkd_loss": 0.01748609459027648 + }, + { + "epoch": 1.624790619765494, + "flow_matching_loss": 0.02682869080454111, + "grad_norm": 0.27871882915496826, + "learning_rate": 2.8335698725124843e-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": 0.015229499991983175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0072166129481047395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037675613351166246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02682869080454111, + "weighted_rkd_loss": 0.018837806675583123 + }, + { + "epoch": 1.641541038525963, + "flow_matching_loss": 0.027610163576900958, + "grad_norm": 0.27080997824668884, + "learning_rate": 2.8289988212339422e-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": 0.014654815848916768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007433690736070276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03674332220107317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027610163576900958, + "weighted_rkd_loss": 0.018371661100536583 + }, + { + "epoch": 1.658291457286432, + "flow_matching_loss": 0.026477544382214547, + "grad_norm": 0.29053348302841187, + "learning_rate": 2.8243696352415997e-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": 0.014315103739500045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006944430945441127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035574638098478314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026477544382214547, + "weighted_rkd_loss": 0.017787319049239157 + }, + { + "epoch": 1.675041876046901, + "flow_matching_loss": 0.03149757422506809, + "grad_norm": 0.24316534399986267, + "learning_rate": 2.8196825170310495e-05, + "loss": 0.0489, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013985770475119352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006879748497158289, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034851289354264736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03149757422506809, + "weighted_rkd_loss": 0.017425644677132368 + }, + { + "epoch": 1.69179229480737, + "flow_matching_loss": 0.028719986975193023, + "grad_norm": 0.2793906033039093, + "learning_rate": 2.8149376716320256e-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": 0.014713972900062799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007857491169124842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03728543724864721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028719986975193023, + "weighted_rkd_loss": 0.018642718624323606 + }, + { + "epoch": 1.708542713567839, + "flow_matching_loss": 0.02652828861027956, + "grad_norm": 0.24760262668132782, + "learning_rate": 2.8101353065994373e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015948845073580742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008608205150812865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04050589501857758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02652828861027956, + "weighted_rkd_loss": 0.02025294750928879 + }, + { + "epoch": 1.725293132328308, + "flow_matching_loss": 0.02748129330575466, + "grad_norm": 0.2966518998146057, + "learning_rate": 2.8052756320042885e-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": 0.013009252492338419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00621428107842803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03223278634250164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02748129330575466, + "weighted_rkd_loss": 0.01611639317125082 + }, + { + "epoch": 1.742043551088777, + "flow_matching_loss": 0.02748612593859434, + "grad_norm": 0.2400917261838913, + "learning_rate": 2.8003588604244893e-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": 0.012675477657467127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006017118878662586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031368074007332325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02748612593859434, + "weighted_rkd_loss": 0.015684037003666162 + }, + { + "epoch": 1.758793969849246, + "flow_matching_loss": 0.0283775020390749, + "grad_norm": 0.23072923719882965, + "learning_rate": 2.7953852069355565e-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": 0.012886785715818406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007881310116499663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03365488145500421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0283775020390749, + "weighted_rkd_loss": 0.016827440727502106 + }, + { + "epoch": 1.775544388609715, + "flow_matching_loss": 0.028830614313483237, + "grad_norm": 0.19938981533050537, + "learning_rate": 2.790354889101206e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011350309383124113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005617117416113615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028317735902965067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028830614313483237, + "weighted_rkd_loss": 0.014158867951482534 + }, + { + "epoch": 1.792294807370184, + "flow_matching_loss": 0.02743749525398016, + "grad_norm": 0.20770491659641266, + "learning_rate": 2.7852681269638366e-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": 0.013101563602685929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006368691730313003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03257181905210018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02743749525398016, + "weighted_rkd_loss": 0.01628590952605009 + }, + { + "epoch": 1.809045226130653, + "flow_matching_loss": 0.025509939529001713, + "grad_norm": 0.23055991530418396, + "learning_rate": 2.780125143034902e-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": 0.01217031548731029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005663800402544439, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030004431121051313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025509939529001713, + "weighted_rkd_loss": 0.015002215560525656 + }, + { + "epoch": 1.8257956448911221, + "flow_matching_loss": 0.027603413537144662, + "grad_norm": 0.24784690141677856, + "learning_rate": 2.7749261622851812e-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": 0.012131767626851798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005731426505371928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02999496180564165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027603413537144662, + "weighted_rkd_loss": 0.014997480902820826 + }, + { + "epoch": 1.8425460636515911, + "flow_matching_loss": 0.02833728678524494, + "grad_norm": 0.2339480221271515, + "learning_rate": 2.7696714121349335e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011987958382815123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005508459242992103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029484376311302185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02833728678524494, + "weighted_rkd_loss": 0.014742188155651093 + }, + { + "epoch": 1.8592964824120601, + "flow_matching_loss": 0.028176490403711795, + "grad_norm": 0.21943725645542145, + "learning_rate": 2.764361122443954e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012409442663192749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006674071168527007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03149295635521412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028176490403711795, + "weighted_rkd_loss": 0.01574647817760706 + }, + { + "epoch": 1.8760469011725294, + "flow_matching_loss": 0.02386368773877621, + "grad_norm": 0.28784745931625366, + "learning_rate": 2.758995525501517e-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": 0.012525459285825491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006127721210941672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031178639829158784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02386368773877621, + "weighted_rkd_loss": 0.015589319914579392 + }, + { + "epoch": 1.8927973199329984, + "flow_matching_loss": 0.026401301100850105, + "grad_norm": 0.302274614572525, + "learning_rate": 2.753574856016215e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012502617854624986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005733438022434712, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03073867317289114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026401301100850105, + "weighted_rkd_loss": 0.01536933658644557 + }, + { + "epoch": 1.9095477386934674, + "flow_matching_loss": 0.025203173607587816, + "grad_norm": 0.22959397733211517, + "learning_rate": 2.7480993511056913e-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": 0.012079661898314952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005447410838678479, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029606734588742255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025203173607587816, + "weighted_rkd_loss": 0.014803367294371128 + }, + { + "epoch": 1.9262981574539364, + "flow_matching_loss": 0.02645403500646353, + "grad_norm": 0.20805855095386505, + "learning_rate": 2.7425692502862684e-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": 0.012591348122805357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005284116393886507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030466812662780285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02645403500646353, + "weighted_rkd_loss": 0.015233406331390142 + }, + { + "epoch": 1.9430485762144054, + "flow_matching_loss": 0.028687219321727752, + "grad_norm": 0.23209843039512634, + "learning_rate": 2.7369847954624728e-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": 0.01346703628078103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005913856625556946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032847929000854495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028687219321727752, + "weighted_rkd_loss": 0.016423964500427247 + }, + { + "epoch": 1.9597989949748744, + "flow_matching_loss": 0.0267184529453516, + "grad_norm": 0.20264476537704468, + "learning_rate": 2.731346230916448e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012482394743710757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005907515040598809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030872304178774356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0267184529453516, + "weighted_rkd_loss": 0.015436152089387178 + }, + { + "epoch": 1.9765494137353434, + "flow_matching_loss": 0.02526174336671829, + "grad_norm": 0.2656524181365967, + "learning_rate": 2.7256538032972746e-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": 0.011073235142976045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004903874779120088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02705034501850605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02526174336671829, + "weighted_rkd_loss": 0.013525172509253025 + }, + { + "epoch": 1.9932998324958124, + "flow_matching_loss": 0.024283120594918727, + "grad_norm": 0.18750469386577606, + "learning_rate": 2.7199077616101773e-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": 0.010565282870084047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005318943597376347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026449509151279926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024283120594918727, + "weighted_rkd_loss": 0.013224754575639963 + }, + { + "epoch": 2.0100502512562812, + "flow_matching_loss": 0.023823299817740916, + "grad_norm": 0.28387999534606934, + "learning_rate": 2.714108357205634e-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": 0.01328891459852457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061988724861294035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032776701636612414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023823299817740916, + "weighted_rkd_loss": 0.016388350818306207 + }, + { + "epoch": 2.0268006700167502, + "flow_matching_loss": 0.027536653354763983, + "grad_norm": 0.23609916865825653, + "learning_rate": 2.7082558437683803e-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": 0.010250479681417346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046427268767729405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025143686309456827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027536653354763983, + "weighted_rkd_loss": 0.012571843154728413 + }, + { + "epoch": 2.0435510887772192, + "flow_matching_loss": 0.025359765440225602, + "grad_norm": 0.2570204436779022, + "learning_rate": 2.702350477306315e-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": 0.012231682194396853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005420093261636794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02988345790654421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025359765440225602, + "weighted_rkd_loss": 0.014941728953272105 + }, + { + "epoch": 2.0603015075376883, + "flow_matching_loss": 0.02634656634181738, + "grad_norm": 0.22433871030807495, + "learning_rate": 2.6963925161392963e-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": 0.010704432427883149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047349676489830015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026143832691013814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02634656634181738, + "weighted_rkd_loss": 0.013071916345506907 + }, + { + "epoch": 2.0770519262981573, + "flow_matching_loss": 0.02522501554340124, + "grad_norm": 0.2080315798521042, + "learning_rate": 2.6903822208878476e-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": 0.010163325816392899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004687550477683544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02501420211046934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02522501554340124, + "weighted_rkd_loss": 0.01250710105523467 + }, + { + "epoch": 2.0938023450586263, + "flow_matching_loss": 0.02463902086019516, + "grad_norm": 0.25818106532096863, + "learning_rate": 2.6843198544617535e-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": 0.012039071694016457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006278580939397216, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030356723815202713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02463902086019516, + "weighted_rkd_loss": 0.015178361907601357 + }, + { + "epoch": 2.1105527638190953, + "flow_matching_loss": 0.027839330770075322, + "grad_norm": 0.2640082538127899, + "learning_rate": 2.6782056820485612e-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": 0.012098926212638616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005462716426700353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029660568572580814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027839330770075322, + "weighted_rkd_loss": 0.014830284286290407 + }, + { + "epoch": 2.1273031825795643, + "flow_matching_loss": 0.028276474215090275, + "grad_norm": 0.295583575963974, + "learning_rate": 2.672039971101979e-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": 0.013212186470627784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006356807774864137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032781181018799545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028276474215090275, + "weighted_rkd_loss": 0.016390590509399772 + }, + { + "epoch": 2.1440536013400333, + "flow_matching_loss": 0.026197852566838264, + "grad_norm": 0.20729024708271027, + "learning_rate": 2.665822991330178e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01121538388542831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005526613956317305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027957382053136824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026197852566838264, + "weighted_rkd_loss": 0.013978691026568412 + }, + { + "epoch": 2.1608040201005023, + "flow_matching_loss": 0.02445336002856493, + "grad_norm": 0.19470828771591187, + "learning_rate": 2.6595550146839934e-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": 0.011698291730135679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005225977627560496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02862256094813347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02445336002856493, + "weighted_rkd_loss": 0.014311280474066735 + }, + { + "epoch": 2.1775544388609713, + "flow_matching_loss": 0.02761701140552759, + "grad_norm": 0.29102823138237, + "learning_rate": 2.6532363153450287e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012762863840907811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00593646764755249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031462194956839086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02761701140552759, + "weighted_rkd_loss": 0.015731097478419543 + }, + { + "epoch": 2.1943048576214403, + "flow_matching_loss": 0.023142579942941666, + "grad_norm": 0.2832508087158203, + "learning_rate": 2.6468671697136634e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01395815759897232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007116776378825307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03503309153020382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023142579942941666, + "weighted_rkd_loss": 0.01751654576510191 + }, + { + "epoch": 2.2110552763819094, + "flow_matching_loss": 0.024112015962600708, + "grad_norm": 0.22076784074306488, + "learning_rate": 2.640447856396961e-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": 0.010843491833657026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005128723871894181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026815707236528395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024112015962600708, + "weighted_rkd_loss": 0.013407853618264198 + }, + { + "epoch": 2.2278056951423784, + "flow_matching_loss": 0.02480712588876486, + "grad_norm": 0.24121958017349243, + "learning_rate": 2.6339786561964805e-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": 0.010331244813278318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004812800046056509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025475289672613144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02480712588876486, + "weighted_rkd_loss": 0.012737644836306572 + }, + { + "epoch": 2.2445561139028474, + "flow_matching_loss": 0.02151898145675659, + "grad_norm": 0.2851496636867523, + "learning_rate": 2.6274598520959966e-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": 0.011607506917789579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005171733908355236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028386748023331167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02151898145675659, + "weighted_rkd_loss": 0.014193374011665583 + }, + { + "epoch": 2.2613065326633164, + "flow_matching_loss": 0.02450380623340607, + "grad_norm": 0.19669336080551147, + "learning_rate": 2.6208917292491185e-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": 0.011534726107493043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005064515001140535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02813396751880646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02450380623340607, + "weighted_rkd_loss": 0.01406698375940323 + }, + { + "epoch": 2.2780569514237854, + "flow_matching_loss": 0.0243028256110847, + "grad_norm": 0.22813156247138977, + "learning_rate": 2.6142745749668176e-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": 0.008905993402004242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004185553826391697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021997540444135665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0243028256110847, + "weighted_rkd_loss": 0.010998770222067833 + }, + { + "epoch": 2.2948073701842544, + "flow_matching_loss": 0.023458981513977052, + "grad_norm": 0.1727350801229477, + "learning_rate": 2.607608678704859e-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": 0.010793544258922339, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005353818205185235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02694090660661459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023458981513977052, + "weighted_rkd_loss": 0.013470453303307294 + }, + { + "epoch": 2.3115577889447234, + "flow_matching_loss": 0.02446927558630705, + "grad_norm": 0.20639468729496002, + "learning_rate": 2.600894332051139e-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": 0.01072792410850525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047226678114384415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026178515702486038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02446927558630705, + "weighted_rkd_loss": 0.013089257851243019 + }, + { + "epoch": 2.3283082077051924, + "flow_matching_loss": 0.02812897562980652, + "grad_norm": 0.23591165244579315, + "learning_rate": 2.594131828712932e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010303571028634906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005087628937326371, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02569477092474699, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02812897562980652, + "weighted_rkd_loss": 0.012847385462373495 + }, + { + "epoch": 2.3450586264656614, + "flow_matching_loss": 0.023055185005068778, + "grad_norm": 0.2851240336894989, + "learning_rate": 2.5873214645040414e-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": 0.011493522580713034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053280909080058335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028315136022865773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023055185005068778, + "weighted_rkd_loss": 0.014157568011432886 + }, + { + "epoch": 2.3618090452261304, + "flow_matching_loss": 0.023752907849848272, + "grad_norm": 0.2843516170978546, + "learning_rate": 2.5804635373318604e-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": 0.012701206561177968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00591087422799319, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03131328746676445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023752907849848272, + "weighted_rkd_loss": 0.015656643733382224 + }, + { + "epoch": 2.3785594639865995, + "flow_matching_loss": 0.024771238677203655, + "grad_norm": 0.23561623692512512, + "learning_rate": 2.57355834718434e-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": 0.008858569525182247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004208746063522994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02192588523030281, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024771238677203655, + "weighted_rkd_loss": 0.010962942615151405 + }, + { + "epoch": 2.3953098827470685, + "flow_matching_loss": 0.024008012004196645, + "grad_norm": 0.20277637243270874, + "learning_rate": 2.566606196116867e-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": 0.01078841257840395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050114454235881565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0265882708132267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024008012004196645, + "weighted_rkd_loss": 0.01329413540661335 + }, + { + "epoch": 2.4120603015075375, + "flow_matching_loss": 0.02423899285495281, + "grad_norm": 0.21418488025665283, + "learning_rate": 2.559607388239051e-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": 0.010564777301624417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005194381531327963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02632393632084131, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02423899285495281, + "weighted_rkd_loss": 0.013161968160420656 + }, + { + "epoch": 2.4288107202680065, + "flow_matching_loss": 0.021590460650622846, + "grad_norm": 0.23149818181991577, + "learning_rate": 2.552562229701422e-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": 0.010688819037750364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005056107765994966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026433745771646498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021590460650622846, + "weighted_rkd_loss": 0.013216872885823249 + }, + { + "epoch": 2.4455611390284755, + "flow_matching_loss": 0.0240889685228467, + "grad_norm": 0.1726461946964264, + "learning_rate": 2.545471028682036e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009562594909220934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043014321010559796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023426621593534945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0240889685228467, + "weighted_rkd_loss": 0.011713310796767473 + }, + { + "epoch": 2.4623115577889445, + "flow_matching_loss": 0.025627955421805383, + "grad_norm": 0.2549595236778259, + "learning_rate": 2.538334095373e-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": 0.010514754336327314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004727694508619607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025757203251123427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025627955421805383, + "weighted_rkd_loss": 0.012878601625561713 + }, + { + "epoch": 2.4790619765494135, + "flow_matching_loss": 0.024999511800706385, + "grad_norm": 0.20995697379112244, + "learning_rate": 2.531151741966896e-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": 0.00883528389967978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004498904151841998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02216947190463543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024999511800706385, + "weighted_rkd_loss": 0.011084735952317714 + }, + { + "epoch": 2.4958123953098825, + "flow_matching_loss": 0.024381516501307487, + "grad_norm": 0.22541256248950958, + "learning_rate": 2.5239242826431302e-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": 0.009589857840910553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004377200687304139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023556916601955892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024381516501307487, + "weighted_rkd_loss": 0.011778458300977946 + }, + { + "epoch": 2.5125628140703515, + "flow_matching_loss": 0.02520757894963026, + "grad_norm": 0.2073884904384613, + "learning_rate": 2.5166520335541865e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010272404458373786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051223641028627755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02566717304289341, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02520757894963026, + "weighted_rkd_loss": 0.012833586521446705 + }, + { + "epoch": 2.5293132328308205, + "flow_matching_loss": 0.02480332348495722, + "grad_norm": 0.22228868305683136, + "learning_rate": 2.5093353128117996e-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": 0.008547855308279395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004206642787903547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021302353404462338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02480332348495722, + "weighted_rkd_loss": 0.010651176702231169 + }, + { + "epoch": 2.5460636515912896, + "flow_matching_loss": 0.023200121708214283, + "grad_norm": 0.1978033035993576, + "learning_rate": 2.501974440473037e-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": 0.010471200989559292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004926082771271467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025868484750390053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023200121708214283, + "weighted_rkd_loss": 0.012934242375195026 + }, + { + "epoch": 2.5628140703517586, + "flow_matching_loss": 0.02555184094235301, + "grad_norm": 0.22602304816246033, + "learning_rate": 2.4945697385263015e-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": 0.008761562407016754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039977532578632236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021520878095179795, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02555184094235301, + "weighted_rkd_loss": 0.010760439047589897 + }, + { + "epoch": 2.5795644891122276, + "flow_matching_loss": 0.021686122007668018, + "grad_norm": 0.1781231164932251, + "learning_rate": 2.487121530877243e-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": 0.010080217849463224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00466340493876487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024823840521275998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021686122007668018, + "weighted_rkd_loss": 0.012411920260637999 + }, + { + "epoch": 2.5963149078726966, + "flow_matching_loss": 0.022586009837687014, + "grad_norm": 0.21403099596500397, + "learning_rate": 2.4796301433345943e-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": 0.010508734360337257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005213955650106072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02623142395168543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022586009837687014, + "weighted_rkd_loss": 0.013115711975842714 + }, + { + "epoch": 2.6130653266331656, + "flow_matching_loss": 0.023120622523128985, + "grad_norm": 0.18488554656505585, + "learning_rate": 2.4720959035959148e-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": 0.008027114858850836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003637227532453835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019691457133740188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023120622523128985, + "weighted_rkd_loss": 0.009845728566870094 + }, + { + "epoch": 2.6298157453936346, + "flow_matching_loss": 0.02213309984654188, + "grad_norm": 0.2590029537677765, + "learning_rate": 2.464519141233257e-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": 0.008783003455027938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004119690437801182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021685697697103025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02213309984654188, + "weighted_rkd_loss": 0.010842848848551512 + }, + { + "epoch": 2.6465661641541036, + "flow_matching_loss": 0.023288248293101787, + "grad_norm": 0.20496879518032074, + "learning_rate": 2.456900187678754e-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": 0.009698561951518058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004631320876069367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024028444662690163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023288248293101787, + "weighted_rkd_loss": 0.012014222331345081 + }, + { + "epoch": 2.6633165829145726, + "flow_matching_loss": 0.02162441601976752, + "grad_norm": 0.1816551685333252, + "learning_rate": 2.4492393762101147e-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": 0.009497049171477557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004680510167963803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023674608673900366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02162441601976752, + "weighted_rkd_loss": 0.011837304336950183 + }, + { + "epoch": 2.6800670016750416, + "flow_matching_loss": 0.02216140553355217, + "grad_norm": 0.1991337537765503, + "learning_rate": 2.4415370419360508e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009966886276379228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004221933707594871, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024155706353485584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02216140553355217, + "weighted_rkd_loss": 0.012077853176742792 + }, + { + "epoch": 2.6968174204355106, + "flow_matching_loss": 0.02192650279030204, + "grad_norm": 0.2767055034637451, + "learning_rate": 2.4337935217816157e-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": 0.010798801574856043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047045488143339755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02630215175449848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02192650279030204, + "weighted_rkd_loss": 0.01315107587724924 + }, + { + "epoch": 2.7135678391959797, + "flow_matching_loss": 0.02603024896234274, + "grad_norm": 0.20454247295856476, + "learning_rate": 2.4260091544734665e-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": 0.010186881572008134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004692598734982312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025066361762583256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02603024896234274, + "weighted_rkd_loss": 0.012533180881291628 + }, + { + "epoch": 2.7303182579564487, + "flow_matching_loss": 0.025671250931918622, + "grad_norm": 0.20730584859848022, + "learning_rate": 2.4181842805250466e-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": 0.008627374796196819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042751003755256535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02152985017746687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025671250931918622, + "weighted_rkd_loss": 0.010764925088733435 + }, + { + "epoch": 2.7470686767169177, + "flow_matching_loss": 0.02359693106263876, + "grad_norm": 0.188655287027359, + "learning_rate": 2.4103192422216923e-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": 0.010495147341862321, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005308826128020883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02629912067204714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02359693106263876, + "weighted_rkd_loss": 0.01314956033602357 + }, + { + "epoch": 2.7638190954773867, + "flow_matching_loss": 0.023337428458034992, + "grad_norm": 0.23929449915885925, + "learning_rate": 2.402414383605658e-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": 0.010052195470780134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004961214587092399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025065605714917182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023337428458034992, + "weighted_rkd_loss": 0.012532802857458591 + }, + { + "epoch": 2.7805695142378557, + "flow_matching_loss": 0.02219377104192972, + "grad_norm": 0.2548065185546875, + "learning_rate": 2.3944700504610674e-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": 0.010901725804433226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045725111151114104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026375962886959314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02219377104192972, + "weighted_rkd_loss": 0.013187981443479657 + }, + { + "epoch": 2.7973199329983247, + "flow_matching_loss": 0.021554847713559866, + "grad_norm": 0.19051788747310638, + "learning_rate": 2.3864865902987883e-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": 0.009427066240459681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004298139782622457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023152272403240203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021554847713559866, + "weighted_rkd_loss": 0.011576136201620102 + }, + { + "epoch": 2.8140703517587937, + "flow_matching_loss": 0.025536979362368584, + "grad_norm": 0.21837979555130005, + "learning_rate": 2.378464352341231e-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": 0.010489001916721464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051647582557052376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026142761670053007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025536979362368584, + "weighted_rkd_loss": 0.013071380835026503 + }, + { + "epoch": 2.8308207705192627, + "flow_matching_loss": 0.020257471688091756, + "grad_norm": 0.21359702944755554, + "learning_rate": 2.3704036875070728e-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": 0.008906964818015695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004341627052053809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022155556827783585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020257471688091756, + "weighted_rkd_loss": 0.011077778413891792 + }, + { + "epoch": 2.8475711892797317, + "flow_matching_loss": 0.022737276274710894, + "grad_norm": 0.18399080634117126, + "learning_rate": 2.362304948395906e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009278100775554776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00438445636536926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022940658032894135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022737276274710894, + "weighted_rkd_loss": 0.011470329016447067 + }, + { + "epoch": 2.8643216080402008, + "flow_matching_loss": 0.02296969071030617, + "grad_norm": 0.2073369175195694, + "learning_rate": 2.3541684892728157e-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": 0.008697510184720158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004495504498481751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02189052477478981, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02296969071030617, + "weighted_rkd_loss": 0.010945262387394905 + }, + { + "epoch": 2.8810720268006698, + "flow_matching_loss": 0.02656156402081251, + "grad_norm": 0.2169615775346756, + "learning_rate": 2.345994666052883e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009356808802112936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004665453173220158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02337907087057829, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02656156402081251, + "weighted_rkd_loss": 0.011689535435289144 + }, + { + "epoch": 2.8978224455611388, + "flow_matching_loss": 0.02025458849966526, + "grad_norm": 0.19369857013225555, + "learning_rate": 2.337783836285614e-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": 0.009551549004390836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004140716185793281, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023243813868612052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02025458849966526, + "weighted_rkd_loss": 0.011621906934306026 + }, + { + "epoch": 2.914572864321608, + "flow_matching_loss": 0.020585576817393304, + "grad_norm": 0.2022498995065689, + "learning_rate": 2.3295363591393026e-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": 0.008104678569361567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003487278800457716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019696636125445366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020585576817393304, + "weighted_rkd_loss": 0.009848318062722683 + }, + { + "epoch": 2.931323283082077, + "flow_matching_loss": 0.02213856503367424, + "grad_norm": 0.22983857989311218, + "learning_rate": 2.3212525953853174e-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": 0.009665360115468502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004478284833021462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023809005320072175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02213856503367424, + "weighted_rkd_loss": 0.011904502660036087 + }, + { + "epoch": 2.948073701842546, + "flow_matching_loss": 0.02260482348501682, + "grad_norm": 0.2563987076282501, + "learning_rate": 2.3129329073823206e-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": 0.008846775069832802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004327321448363364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02202087165787816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02260482348501682, + "weighted_rkd_loss": 0.01101043582893908 + }, + { + "epoch": 2.964824120603015, + "flow_matching_loss": 0.024081751145422457, + "grad_norm": 0.27132245898246765, + "learning_rate": 2.3045776590604175e-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": 0.011177179915830494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004908815934322774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027263175696134567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024081751145422457, + "weighted_rkd_loss": 0.013631587848067284 + }, + { + "epoch": 2.981574539363484, + "flow_matching_loss": 0.02145145069807768, + "grad_norm": 0.23230737447738647, + "learning_rate": 2.296187215905237e-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": 0.00857270359992981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039184731664136056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021063880529254675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02145145069807768, + "weighted_rkd_loss": 0.010531940264627337 + }, + { + "epoch": 2.998324958123953, + "flow_matching_loss": 0.022610493656247854, + "grad_norm": 0.21160763502120972, + "learning_rate": 2.2877619449419437e-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": 0.00877288361079991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040158325573429465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02156159970909357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022610493656247854, + "weighted_rkd_loss": 0.010780799854546785 + }, + { + "epoch": 3.0150753768844223, + "flow_matching_loss": 0.01961312396451831, + "grad_norm": 0.21592052280902863, + "learning_rate": 2.2793022147191843e-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": 0.008552303677424789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004099966096691787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021204573288559913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01961312396451831, + "weighted_rkd_loss": 0.010602286644279956 + }, + { + "epoch": 3.0318257956448913, + "flow_matching_loss": 0.021133813075721264, + "grad_norm": 0.2408725470304489, + "learning_rate": 2.2708083952929634e-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": 0.009924392122775317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004459633771330118, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024308417923748493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021133813075721264, + "weighted_rkd_loss": 0.012154208961874247 + }, + { + "epoch": 3.0485762144053603, + "flow_matching_loss": 0.021858297660946845, + "grad_norm": 0.21889373660087585, + "learning_rate": 2.2622808582104594e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007935315230861306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036201345268636943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019490764942020178, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021858297660946845, + "weighted_rkd_loss": 0.009745382471010089 + }, + { + "epoch": 3.0653266331658293, + "flow_matching_loss": 0.02066167173907161, + "grad_norm": 0.2489049881696701, + "learning_rate": 2.2537199764937703e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008639211347326637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042247757315635685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021503198333084582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02066167173907161, + "weighted_rkd_loss": 0.010751599166542291 + }, + { + "epoch": 3.0820770519262983, + "flow_matching_loss": 0.021619039587676525, + "grad_norm": 0.18713706731796265, + "learning_rate": 2.245126124623595e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00882501546293497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004201194737106562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021851225569844247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021619039587676525, + "weighted_rkd_loss": 0.010925612784922123 + }, + { + "epoch": 3.0988274706867673, + "flow_matching_loss": 0.0222555180080235, + "grad_norm": 0.16398872435092926, + "learning_rate": 2.2364996785228547e-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": 0.008472857438027859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038555797887966036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020801294595003128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0222555180080235, + "weighted_rkd_loss": 0.010400647297501564 + }, + { + "epoch": 3.1155778894472363, + "flow_matching_loss": 0.02233735201880336, + "grad_norm": 0.195217564702034, + "learning_rate": 2.227841015540248e-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": 0.00930324881337583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003956924844533205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022563422471284865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02233735201880336, + "weighted_rkd_loss": 0.011281711235642432 + }, + { + "epoch": 3.1323283082077054, + "flow_matching_loss": 0.021456127893179654, + "grad_norm": 0.24798332154750824, + "learning_rate": 2.2191505144337433e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00950444871559739, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004383202618919313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023392099887132645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021456127893179654, + "weighted_rkd_loss": 0.011696049943566322 + }, + { + "epoch": 3.1490787269681744, + "flow_matching_loss": 0.020449751988053322, + "grad_norm": 0.21128728985786438, + "learning_rate": 2.210428555354013e-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": 0.007659428799524903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003622169536538422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018941027391701937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020449751988053322, + "weighted_rkd_loss": 0.009470513695850969 + }, + { + "epoch": 3.1658291457286434, + "flow_matching_loss": 0.020440852362662555, + "grad_norm": 0.19786065816879272, + "learning_rate": 2.201675519827802e-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": 0.00879075680859387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00397156581748277, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02155307959765196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020440852362662555, + "weighted_rkd_loss": 0.01077653979882598 + }, + { + "epoch": 3.1825795644891124, + "flow_matching_loss": 0.021140540950000286, + "grad_norm": 0.17299114167690277, + "learning_rate": 2.1928917907412403e-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": 0.008886780729517341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004150486085563898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021924047172069548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021140540950000286, + "weighted_rkd_loss": 0.010962023586034774 + }, + { + "epoch": 3.1993299832495814, + "flow_matching_loss": 0.021827292442321778, + "grad_norm": 0.29607075452804565, + "learning_rate": 2.1840777523230937e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008809508057311177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038985042599961163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021517520304769278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021827292442321778, + "weighted_rkd_loss": 0.010758760152384639 + }, + { + "epoch": 3.2160804020100504, + "flow_matching_loss": 0.02189473919570446, + "grad_norm": 0.21935483813285828, + "learning_rate": 2.1752337901279552e-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": 0.008246973948553204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004177061165682972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02067100889980793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02189473919570446, + "weighted_rkd_loss": 0.010335504449903965 + }, + { + "epoch": 3.2328308207705194, + "flow_matching_loss": 0.02115060854703188, + "grad_norm": 0.21308168768882751, + "learning_rate": 2.166360291019383e-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": 0.008105326350778342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034979252610355615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01970857810229063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02115060854703188, + "weighted_rkd_loss": 0.009854289051145316 + }, + { + "epoch": 3.2495812395309884, + "flow_matching_loss": 0.019409695640206337, + "grad_norm": 0.2219625562429428, + "learning_rate": 2.157457643152973e-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": 0.007863681856542825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037062383722513913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01943360213190317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019409695640206337, + "weighted_rkd_loss": 0.009716801065951585 + }, + { + "epoch": 3.2663316582914574, + "flow_matching_loss": 0.02243496719747782, + "grad_norm": 0.17238378524780273, + "learning_rate": 2.148526235959385e-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": 0.008457799488678575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003480206034146249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020395804755389692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02243496719747782, + "weighted_rkd_loss": 0.010197902377694846 + }, + { + "epoch": 3.2830820770519265, + "flow_matching_loss": 0.021297021582722665, + "grad_norm": 0.21285296976566315, + "learning_rate": 2.1395664601273016e-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": 0.008106729062274099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003673051320947707, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019886509422212838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021297021582722665, + "weighted_rkd_loss": 0.009943254711106419 + }, + { + "epoch": 3.2998324958123955, + "flow_matching_loss": 0.02030411781743169, + "grad_norm": 0.22372476756572723, + "learning_rate": 2.130578707586345e-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": 0.008725496334955096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004435207205824554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021886199712753296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02030411781743169, + "weighted_rkd_loss": 0.010943099856376648 + }, + { + "epoch": 3.3165829145728645, + "flow_matching_loss": 0.021738264430314302, + "grad_norm": 0.20458626747131348, + "learning_rate": 2.1215633714899264e-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": 0.008855098765343428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004159924946725369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021870122756808996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021738264430314302, + "weighted_rkd_loss": 0.010935061378404498 + }, + { + "epoch": 3.3333333333333335, + "flow_matching_loss": 0.022326320968568326, + "grad_norm": 0.20455840229988098, + "learning_rate": 2.112520846198053e-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": 0.008037483179941773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037062566028907896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019781222846359015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022326320968568326, + "weighted_rkd_loss": 0.009890611423179508 + }, + { + "epoch": 3.3500837520938025, + "flow_matching_loss": 0.0228325380012393, + "grad_norm": 0.2642732560634613, + "learning_rate": 2.103451527260074e-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": 0.007869466673582792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033093452220782638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019048278499394656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0228325380012393, + "weighted_rkd_loss": 0.009524139249697328 + }, + { + "epoch": 3.3668341708542715, + "flow_matching_loss": 0.019583611376583578, + "grad_norm": 0.2518172860145569, + "learning_rate": 2.094355811397381e-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": 0.008222047938033938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033477101242169738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0197918058373034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019583611376583578, + "weighted_rkd_loss": 0.0098959029186517 + }, + { + "epoch": 3.3835845896147405, + "flow_matching_loss": 0.024029638990759848, + "grad_norm": 0.19392850995063782, + "learning_rate": 2.0852340964860522e-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": 0.008880104077979923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004434320167638362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022194528207182884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024029638990759848, + "weighted_rkd_loss": 0.011097264103591442 + }, + { + "epoch": 3.4003350083752095, + "flow_matching_loss": 0.02049247771501541, + "grad_norm": 0.1794871985912323, + "learning_rate": 2.076086781539447e-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": 0.007119336817413569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028993004932999613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01713797440752387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02049247771501541, + "weighted_rkd_loss": 0.008568987203761935 + }, + { + "epoch": 3.4170854271356785, + "flow_matching_loss": 0.017801507376134396, + "grad_norm": 0.20647704601287842, + "learning_rate": 2.066914266690755e-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": 0.008224464533850551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00377469330560416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02022362258285284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017801507376134396, + "weighted_rkd_loss": 0.01011181129142642 + }, + { + "epoch": 3.4338358458961475, + "flow_matching_loss": 0.019913967046886683, + "grad_norm": 0.1910584270954132, + "learning_rate": 2.0577169531754906e-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": 0.008146553998813033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033962710993364455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019689378887414934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019913967046886683, + "weighted_rkd_loss": 0.009844689443707467 + }, + { + "epoch": 3.4505862646566166, + "flow_matching_loss": 0.022410364355891943, + "grad_norm": 0.15049278736114502, + "learning_rate": 2.048495243313942e-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": 0.007043796870857477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003043292067013681, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017130885925143956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022410364355891943, + "weighted_rkd_loss": 0.008565442962571978 + }, + { + "epoch": 3.4673366834170856, + "flow_matching_loss": 0.019107151962816715, + "grad_norm": 0.1801232546567917, + "learning_rate": 2.039249540493572e-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": 0.006890717707574367, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027747028158046304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016556138079613446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019107151962816715, + "weighted_rkd_loss": 0.008278069039806723 + }, + { + "epoch": 3.4840871021775546, + "flow_matching_loss": 0.01943407179787755, + "grad_norm": 0.21524561941623688, + "learning_rate": 2.0299802491513755e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007901145797222852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037480345461517574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01955032618716359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01943407179787755, + "weighted_rkd_loss": 0.009775163093581795 + }, + { + "epoch": 3.5008375209380236, + "flow_matching_loss": 0.020493100211024285, + "grad_norm": 0.22612208127975464, + "learning_rate": 2.0206877747561828e-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": 0.0077430652920156716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003920659353025257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019406790286302565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020493100211024285, + "weighted_rkd_loss": 0.009703395143151283 + }, + { + "epoch": 3.5175879396984926, + "flow_matching_loss": 0.017947696149349213, + "grad_norm": 0.21178822219371796, + "learning_rate": 2.011372523790927e-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": 0.008696998190134763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004025266319513321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021419262699782848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017947696149349213, + "weighted_rkd_loss": 0.010709631349891424 + }, + { + "epoch": 3.5343383584589616, + "flow_matching_loss": 0.018908549472689627, + "grad_norm": 0.20278148353099823, + "learning_rate": 2.0020349037348643e-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": 0.008784465352073311, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003880465682595968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02144939647987485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018908549472689627, + "weighted_rkd_loss": 0.010724698239937425 + }, + { + "epoch": 3.5510887772194306, + "flow_matching_loss": 0.018494799360632898, + "grad_norm": 0.1947580724954605, + "learning_rate": 1.9926753230457443e-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": 0.009449501568451524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004069603490643203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022968606650829317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018494799360632898, + "weighted_rkd_loss": 0.011484303325414658 + }, + { + "epoch": 3.5678391959798996, + "flow_matching_loss": 0.02321052923798561, + "grad_norm": 0.21995417773723602, + "learning_rate": 1.983294191141948e-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": 0.007246848475188017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003205891884863377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017699589021503925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02321052923798561, + "weighted_rkd_loss": 0.008849794510751963 + }, + { + "epoch": 3.5845896147403686, + "flow_matching_loss": 0.02240783693268895, + "grad_norm": 0.19936057925224304, + "learning_rate": 1.973891918384575e-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": 0.009624690655618906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004620609059929848, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02386999037116766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02240783693268895, + "weighted_rkd_loss": 0.01193499518558383 + }, + { + "epoch": 3.6013400335008376, + "flow_matching_loss": 0.02322760485112667, + "grad_norm": 0.21677754819393158, + "learning_rate": 1.9644689160594945e-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": 0.0078708321088925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003928346477914601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019670010544359684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02322760485112667, + "weighted_rkd_loss": 0.009835005272179842 + }, + { + "epoch": 3.6180904522613067, + "flow_matching_loss": 0.019495286140590908, + "grad_norm": 0.22844098508358002, + "learning_rate": 1.9550255963593553e-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": 0.007603819016367197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035069640493020416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01871460219845176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019495286140590908, + "weighted_rkd_loss": 0.00935730109922588 + }, + { + "epoch": 3.6348408710217757, + "flow_matching_loss": 0.020784722454845906, + "grad_norm": 0.19066329300403595, + "learning_rate": 1.9455623723655524e-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": 0.006900070700794458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033538440940901635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01715398533269763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020784722454845906, + "weighted_rkd_loss": 0.008576992666348815 + }, + { + "epoch": 3.6515912897822447, + "flow_matching_loss": 0.02208636198192835, + "grad_norm": 0.12610961496829987, + "learning_rate": 1.9360796580301606e-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": 0.006914446130394936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028566413442604245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016685533430427314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02208636198192835, + "weighted_rkd_loss": 0.008342766715213657 + }, + { + "epoch": 3.6683417085427137, + "flow_matching_loss": 0.02097361385822296, + "grad_norm": 0.19243593513965607, + "learning_rate": 1.9265778681578245e-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": 0.007681887224316597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037624599412083627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0191262342967093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02097361385822296, + "weighted_rkd_loss": 0.00956311714835465 + }, + { + "epoch": 3.6850921273031827, + "flow_matching_loss": 0.019382779020816088, + "grad_norm": 0.2503208518028259, + "learning_rate": 1.9170574183876143e-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": 0.008374608866870404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038465243531391026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02059574192389846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019382779020816088, + "weighted_rkd_loss": 0.01029787096194923 + }, + { + "epoch": 3.7018425460636517, + "flow_matching_loss": 0.01804492184892297, + "grad_norm": 0.16802993416786194, + "learning_rate": 1.9075187251748452e-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": 0.009310635272413492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043102250900119545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02293149596080184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01804492184892297, + "weighted_rkd_loss": 0.01146574798040092 + }, + { + "epoch": 3.7185929648241207, + "flow_matching_loss": 0.020559789799153804, + "grad_norm": 0.20331931114196777, + "learning_rate": 1.8979622057728596e-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": 0.007381893135607243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030734885251149533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017837275005877017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020559789799153804, + "weighted_rkd_loss": 0.008918637502938509 + }, + { + "epoch": 3.7353433835845897, + "flow_matching_loss": 0.021128038968890907, + "grad_norm": 0.2322036772966385, + "learning_rate": 1.8883882782147748e-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": 0.007587933773174882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003291653376072645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01846752082929015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021128038968890907, + "weighted_rkd_loss": 0.009233760414645076 + }, + { + "epoch": 3.7520938023450587, + "flow_matching_loss": 0.020190212596207856, + "grad_norm": 0.2527245879173279, + "learning_rate": 1.8787973612951967e-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": 0.008083526277914643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038183052791282536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01998535804450512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020190212596207856, + "weighted_rkd_loss": 0.00999267902225256 + }, + { + "epoch": 3.7688442211055277, + "flow_matching_loss": 0.020125599578022957, + "grad_norm": 0.18612495064735413, + "learning_rate": 1.8691898745519028e-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": 0.007300873147323727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002893792022950947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017495538201183082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020125599578022957, + "weighted_rkd_loss": 0.008747769100591541 + }, + { + "epoch": 3.7855946398659968, + "flow_matching_loss": 0.02288515316322446, + "grad_norm": 0.2038274109363556, + "learning_rate": 1.8595662382474863e-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": 0.007445386983454228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003284679213538766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01817545313388109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02288515316322446, + "weighted_rkd_loss": 0.009087726566940546 + }, + { + "epoch": 3.8023450586264658, + "flow_matching_loss": 0.01813176032155752, + "grad_norm": 0.15173381567001343, + "learning_rate": 1.8499268733509766e-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": 0.007210503658279776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003472392586991191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01789340013638139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01813176032155752, + "weighted_rkd_loss": 0.008946700068190695 + }, + { + "epoch": 3.819095477386935, + "flow_matching_loss": 0.021203831862658264, + "grad_norm": 0.21146638691425323, + "learning_rate": 1.84027220151942e-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": 0.0065764828585088255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003243767865933478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016396733839064835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021203831862658264, + "weighted_rkd_loss": 0.008198366919532418 + }, + { + "epoch": 3.835845896147404, + "flow_matching_loss": 0.02068155473098159, + "grad_norm": 0.22891871631145477, + "learning_rate": 1.83060264507944e-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": 0.008209372824057936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003933777997735888, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02035252321511507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02068155473098159, + "weighted_rkd_loss": 0.010176261607557534 + }, + { + "epoch": 3.852596314907873, + "flow_matching_loss": 0.020051562413573264, + "grad_norm": 0.14797094464302063, + "learning_rate": 1.82091862700876e-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": 0.00664952090010047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002952263760380447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01625130567699671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020051562413573264, + "weighted_rkd_loss": 0.008125652838498355 + }, + { + "epoch": 3.869346733668342, + "flow_matching_loss": 0.02093609981238842, + "grad_norm": 0.2215096801519394, + "learning_rate": 1.8112205709177023e-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": 0.00714435912668705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003198638209141791, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017487356252968312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02093609981238842, + "weighted_rkd_loss": 0.008743678126484156 + }, + { + "epoch": 3.886097152428811, + "flow_matching_loss": 0.019973818585276602, + "grad_norm": 0.18120259046554565, + "learning_rate": 1.8015089010306585e-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": 0.0070549966767430305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003187867905944586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01729786125943065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019973818585276602, + "weighted_rkd_loss": 0.008648930629715324 + }, + { + "epoch": 3.90284757118928, + "flow_matching_loss": 0.019720715191215276, + "grad_norm": 0.22681787610054016, + "learning_rate": 1.7917840421675298e-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": 0.007141135260462761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00324872734490782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017530998028814793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019720715191215276, + "weighted_rkd_loss": 0.008765499014407397 + }, + { + "epoch": 3.919597989949749, + "flow_matching_loss": 0.018794780503958462, + "grad_norm": 0.21289195120334625, + "learning_rate": 1.7820464197251474e-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": 0.006023052660748362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029200530145317314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014966158475726844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018794780503958462, + "weighted_rkd_loss": 0.007483079237863422 + }, + { + "epoch": 3.936348408710218, + "flow_matching_loss": 0.02315736413002014, + "grad_norm": 0.22471879422664642, + "learning_rate": 1.7722964596586627e-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": 0.006645664293318987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028370591229759158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01612838786095381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02315736413002014, + "weighted_rkd_loss": 0.008064193930476905 + }, + { + "epoch": 3.953098827470687, + "flow_matching_loss": 0.019443149212747812, + "grad_norm": 0.19242846965789795, + "learning_rate": 1.7625345884629144e-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": 0.007972657866775989, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036050250404514373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019550340808928014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019443149212747812, + "weighted_rkd_loss": 0.009775170404464007 + }, + { + "epoch": 3.969849246231156, + "flow_matching_loss": 0.02088850997388363, + "grad_norm": 0.1732814460992813, + "learning_rate": 1.7527612331537722e-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": 0.006496066693216562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002922359528020024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015914492681622504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02088850997388363, + "weighted_rkd_loss": 0.007957246340811252 + }, + { + "epoch": 3.986599664991625, + "flow_matching_loss": 0.019626106694340707, + "grad_norm": 0.1980288326740265, + "learning_rate": 1.7429768212494593e-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": 0.006757032917812466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003019305528141558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016533371340483428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019626106694340707, + "weighted_rkd_loss": 0.008266685670241714 + }, + { + "epoch": 4.0033500837520934, + "flow_matching_loss": 0.01870811516419053, + "grad_norm": 0.25888529419898987, + "learning_rate": 1.7331817807518497e-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": 0.008479700377210974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036272499710321426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020586650539189576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01870811516419053, + "weighted_rkd_loss": 0.010293325269594788 + }, + { + "epoch": 4.0201005025125625, + "flow_matching_loss": 0.019916127249598504, + "grad_norm": 0.18142595887184143, + "learning_rate": 1.723376540127745e-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": 0.006715080700814724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028278434183448553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016258004866540433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019916127249598504, + "weighted_rkd_loss": 0.008129002433270216 + }, + { + "epoch": 4.0368509212730315, + "flow_matching_loss": 0.01878450671210885, + "grad_norm": 0.23574140667915344, + "learning_rate": 1.7135615282901357e-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": 0.0064266437198966745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003169252898078412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01602254034951329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01878450671210885, + "weighted_rkd_loss": 0.008011270174756646 + }, + { + "epoch": 4.0536013400335005, + "flow_matching_loss": 0.020211434923112392, + "grad_norm": 0.20220205187797546, + "learning_rate": 1.7037371745794358e-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": 0.006330230925232172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002878256363328546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01553871827200055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020211434923112392, + "weighted_rkd_loss": 0.007769359136000275 + }, + { + "epoch": 4.0703517587939695, + "flow_matching_loss": 0.019851094391196967, + "grad_norm": 0.2334161400794983, + "learning_rate": 1.6939039087447026e-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": 0.0066378221847116945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003039022069424391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016314666345715524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019851094391196967, + "weighted_rkd_loss": 0.008157333172857762 + }, + { + "epoch": 4.0871021775544385, + "flow_matching_loss": 0.020226925425231457, + "grad_norm": 0.1534281224012375, + "learning_rate": 1.6840621609248397e-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": 0.007128366222605109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003230903437361121, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017487635742872952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020226925425231457, + "weighted_rkd_loss": 0.008743817871436476 + }, + { + "epoch": 4.1038525963149075, + "flow_matching_loss": 0.019695611950010062, + "grad_norm": 0.2591371536254883, + "learning_rate": 1.6742123616297794e-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": 0.007243454223498702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00313732223585248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017624230589717627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019695611950010062, + "weighted_rkd_loss": 0.008812115294858813 + }, + { + "epoch": 4.1206030150753765, + "flow_matching_loss": 0.018704015389084817, + "grad_norm": 0.13702625036239624, + "learning_rate": 1.664354941721651e-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": 0.007517511304467916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003035618958529085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018070641439408064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018704015389084817, + "weighted_rkd_loss": 0.009035320719704032 + }, + { + "epoch": 4.1373534338358455, + "flow_matching_loss": 0.017411495372653008, + "grad_norm": 0.19919848442077637, + "learning_rate": 1.6544903323959355e-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": 0.007433804357424379, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003234395873732865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01810200437903404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017411495372653008, + "weighted_rkd_loss": 0.00905100218951702 + }, + { + "epoch": 4.1541038525963145, + "flow_matching_loss": 0.02045935718342662, + "grad_norm": 0.21578043699264526, + "learning_rate": 1.6446189651626023e-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": 0.007500055991113186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003137087286449969, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01813719905912876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02045935718342662, + "weighted_rkd_loss": 0.00906859952956438 + }, + { + "epoch": 4.1708542713567835, + "flow_matching_loss": 0.02151382314041257, + "grad_norm": 0.16690842807292938, + "learning_rate": 1.6347412718272316e-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": 0.0064533458091318606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029357226798310877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01584241455420852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02151382314041257, + "weighted_rkd_loss": 0.00792120727710426 + }, + { + "epoch": 4.187604690117253, + "flow_matching_loss": 0.016593755688518286, + "grad_norm": 0.1668848693370819, + "learning_rate": 1.624857684472131e-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": 0.006326651386916638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00287430175812915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015527604427188636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016593755688518286, + "weighted_rkd_loss": 0.007763802213594318 + }, + { + "epoch": 4.204355108877722, + "flow_matching_loss": 0.018163218908011914, + "grad_norm": 0.21210342645645142, + "learning_rate": 1.6149686354374283e-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": 0.006518031936138868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030372726498171685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016073336359113455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018163218908011914, + "weighted_rkd_loss": 0.008036668179556727 + }, + { + "epoch": 4.221105527638191, + "flow_matching_loss": 0.01736935544759035, + "grad_norm": 0.18703977763652802, + "learning_rate": 1.605074557302165e-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": 0.007781515922397375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003655482176691294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01921851420775056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01736935544759035, + "weighted_rkd_loss": 0.00960925710387528 + }, + { + "epoch": 4.23785594639866, + "flow_matching_loss": 0.02085604052990675, + "grad_norm": 0.1353197544813156, + "learning_rate": 1.59517588286537e-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": 0.007296613347716629, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029753129929304123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0175685397349298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02085604052990675, + "weighted_rkd_loss": 0.0087842698674649 + }, + { + "epoch": 4.254606365159129, + "flow_matching_loss": 0.020376010704785584, + "grad_norm": 0.1928730607032776, + "learning_rate": 1.5852730451271324e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007122312113642693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003080862946808338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017325487174093723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020376010704785584, + "weighted_rkd_loss": 0.008662743587046861 + }, + { + "epoch": 4.271356783919598, + "flow_matching_loss": 0.018556658551096915, + "grad_norm": 0.18790344893932343, + "learning_rate": 1.5753664772696546e-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": 0.006062540132552386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028738549328409134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014998935256153345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018556658551096915, + "weighted_rkd_loss": 0.007499467628076672 + }, + { + "epoch": 4.288107202680067, + "flow_matching_loss": 0.018578741513192654, + "grad_norm": 0.20445463061332703, + "learning_rate": 1.565456612638307e-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": 0.006489685783162713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027945277863182126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015773899294435977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018578741513192654, + "weighted_rkd_loss": 0.007886949647217988 + }, + { + "epoch": 4.304857621440536, + "flow_matching_loss": 0.02046471843495965, + "grad_norm": 0.1576392948627472, + "learning_rate": 1.5555438847226732e-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": 0.006184978596866131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025121036567725243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014882060885429382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02046471843495965, + "weighted_rkd_loss": 0.007441030442714691 + }, + { + "epoch": 4.321608040201005, + "flow_matching_loss": 0.02033666577190161, + "grad_norm": 0.19884538650512695, + "learning_rate": 1.545628727137584e-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": 0.006643129396252334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029675282421521842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016253787092864513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02033666577190161, + "weighted_rkd_loss": 0.008126893546432257 + }, + { + "epoch": 4.338358458961474, + "flow_matching_loss": 0.01932725394144654, + "grad_norm": 0.1970246136188507, + "learning_rate": 1.535711573604153e-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": 0.00804528738372028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036893374170176687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019779912009835244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01932725394144654, + "weighted_rkd_loss": 0.009889956004917622 + }, + { + "epoch": 4.355108877721943, + "flow_matching_loss": 0.01995522724464536, + "grad_norm": 0.1644182652235031, + "learning_rate": 1.5257928579308032e-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": 0.006598327262327075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033264036406762897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016523058246821164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01995522724464536, + "weighted_rkd_loss": 0.008261529123410582 + }, + { + "epoch": 4.371859296482412, + "flow_matching_loss": 0.0177932090125978, + "grad_norm": 0.21884872019290924, + "learning_rate": 1.5158730139942902e-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": 0.008010930195450783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033667325507849453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019388592801988125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0177932090125978, + "weighted_rkd_loss": 0.009694296400994062 + }, + { + "epoch": 4.388609715242881, + "flow_matching_loss": 0.019758843723684548, + "grad_norm": 0.1710134893655777, + "learning_rate": 1.5059524757207239e-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": 0.00648601446300745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029278623638674616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015899891406297682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019758843723684548, + "weighted_rkd_loss": 0.007949945703148841 + }, + { + "epoch": 4.40536013400335, + "flow_matching_loss": 0.017516419384628536, + "grad_norm": 0.15637044608592987, + "learning_rate": 1.4960316770665866e-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": 0.006116408109664917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025824596639722587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014815275836735965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017516419384628536, + "weighted_rkd_loss": 0.007407637918367982 + }, + { + "epoch": 4.422110552763819, + "flow_matching_loss": 0.019074277114123107, + "grad_norm": 0.22107526659965515, + "learning_rate": 1.4861110519997495e-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": 0.006211187061853707, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028650935622863472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015287467464804649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019074277114123107, + "weighted_rkd_loss": 0.007643733732402324 + }, + { + "epoch": 4.438860971524288, + "flow_matching_loss": 0.01782265156507492, + "grad_norm": 0.23777632415294647, + "learning_rate": 1.4761910344804913e-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": 0.007094258349388838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028426593402400614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01703117610886693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01782265156507492, + "weighted_rkd_loss": 0.008515588054433466 + }, + { + "epoch": 4.455611390284757, + "flow_matching_loss": 0.020024440810084344, + "grad_norm": 0.1772475689649582, + "learning_rate": 1.466272058442516e-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": 0.006881606904789806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002932384598534554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016695598419755698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020024440810084344, + "weighted_rkd_loss": 0.008347799209877849 + }, + { + "epoch": 4.472361809045226, + "flow_matching_loss": 0.01836460642516613, + "grad_norm": 0.1673913151025772, + "learning_rate": 1.4563545577739676e-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": 0.0067969394847750665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003281451168004423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016875330172479153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01836460642516613, + "weighted_rkd_loss": 0.008437665086239577 + }, + { + "epoch": 4.489112227805695, + "flow_matching_loss": 0.016627813410013913, + "grad_norm": 0.16218630969524384, + "learning_rate": 1.4464389662984543e-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": 0.006768457312136889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003101827809587121, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01663874238729477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016627813410013913, + "weighted_rkd_loss": 0.008319371193647385 + }, + { + "epoch": 4.505862646566165, + "flow_matching_loss": 0.016898567136377097, + "grad_norm": 0.14560790359973907, + "learning_rate": 1.4365257177560684e-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": 0.005781614547595381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002422610018402338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013985839113593102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016898567136377097, + "weighted_rkd_loss": 0.006992919556796551 + }, + { + "epoch": 4.522613065326633, + "flow_matching_loss": 0.01971247997134924, + "grad_norm": 0.14987064898014069, + "learning_rate": 1.426615245784416e-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": 0.007260903157293797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002818218374159187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01734002446755767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01971247997134924, + "weighted_rkd_loss": 0.008670012233778834 + }, + { + "epoch": 4.539363484087103, + "flow_matching_loss": 0.017527775652706624, + "grad_norm": 0.1931697577238083, + "learning_rate": 1.4167079838996468e-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": 0.007004645606502891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003280059481039643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017289350647479294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017527775652706624, + "weighted_rkd_loss": 0.008644675323739647 + }, + { + "epoch": 4.556113902847571, + "flow_matching_loss": 0.018044522684067487, + "grad_norm": 0.18841363489627838, + "learning_rate": 1.4068043654774897e-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": 0.0072848881827667356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003721295588184148, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0182910718023777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018044522684067487, + "weighted_rkd_loss": 0.00914553590118885 + }, + { + "epoch": 4.572864321608041, + "flow_matching_loss": 0.016927159391343594, + "grad_norm": 0.18175353109836578, + "learning_rate": 1.3969048237342992e-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": 0.006235395558178425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026640282245352863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015134819317609072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016927159391343594, + "weighted_rkd_loss": 0.007567409658804536 + }, + { + "epoch": 4.589614740368509, + "flow_matching_loss": 0.020692834444344044, + "grad_norm": 0.17616070806980133, + "learning_rate": 1.3870097917081011e-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": 0.00688162581063807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029462887556292116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01670954031869769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020692834444344044, + "weighted_rkd_loss": 0.008354770159348846 + }, + { + "epoch": 4.606365159128979, + "flow_matching_loss": 0.01656912127509713, + "grad_norm": 0.17389723658561707, + "learning_rate": 1.3771197022396508e-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": 0.0060537876561284065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002805469243321568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014913044683635236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01656912127509713, + "weighted_rkd_loss": 0.007456522341817618 + }, + { + "epoch": 4.623115577889447, + "flow_matching_loss": 0.018882535118609666, + "grad_norm": 0.1804601401090622, + "learning_rate": 1.3672349879535015e-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": 0.0065174739807844165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026487289811484515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015683677047491074, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018882535118609666, + "weighted_rkd_loss": 0.007841838523745537 + }, + { + "epoch": 4.639865996649917, + "flow_matching_loss": 0.018018264696002005, + "grad_norm": 0.18113377690315247, + "learning_rate": 1.3573560812390781e-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": 0.006505385902710259, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00266862609423697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015679398085922003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018018264696002005, + "weighted_rkd_loss": 0.007839699042961001 + }, + { + "epoch": 4.656616415410385, + "flow_matching_loss": 0.017154605966061352, + "grad_norm": 0.22344084084033966, + "learning_rate": 1.3474834142317635e-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": 0.00781757920049131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003766122250817716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019401280861347913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017154605966061352, + "weighted_rkd_loss": 0.009700640430673957 + }, + { + "epoch": 4.673366834170855, + "flow_matching_loss": 0.020884703472256662, + "grad_norm": 0.16884377598762512, + "learning_rate": 1.3376174187939955e-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": 0.00663802158087492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002596668968908489, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015872712060809136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020884703472256662, + "weighted_rkd_loss": 0.007936356030404568 + }, + { + "epoch": 4.690117252931323, + "flow_matching_loss": 0.01987990001216531, + "grad_norm": 0.17503158748149872, + "learning_rate": 1.3277585264963761e-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": 0.006741104461252689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00272078113630414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01620299005880952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01987990001216531, + "weighted_rkd_loss": 0.00810149502940476 + }, + { + "epoch": 4.706867671691793, + "flow_matching_loss": 0.018524412903934716, + "grad_norm": 0.2198275923728943, + "learning_rate": 1.3179071685987928e-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": 0.006492939265444875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026159399072639643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015601818356662989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018524412903934716, + "weighted_rkd_loss": 0.0078009091783314945 + }, + { + "epoch": 4.723618090452261, + "flow_matching_loss": 0.018392641749233007, + "grad_norm": 0.15910054743289948, + "learning_rate": 1.308063776031554e-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": 0.006039766082540155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022717000334523617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01435123234987259, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018392641749233007, + "weighted_rkd_loss": 0.007175616174936295 + }, + { + "epoch": 4.740368509212731, + "flow_matching_loss": 0.016925379633903503, + "grad_norm": 0.1666567623615265, + "learning_rate": 1.2982287793765389e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005827722325921059, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002627475606277585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014282920211553574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016925379633903503, + "weighted_rkd_loss": 0.007141460105776787 + }, + { + "epoch": 4.757118927973199, + "flow_matching_loss": 0.016931863874197005, + "grad_norm": 0.16195693612098694, + "learning_rate": 1.2884026088483616e-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": 0.006095975963398814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00292929713614285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015121249109506607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016931863874197005, + "weighted_rkd_loss": 0.007560624554753303 + }, + { + "epoch": 4.773869346733669, + "flow_matching_loss": 0.02151737492531538, + "grad_norm": 0.16743698716163635, + "learning_rate": 1.2785856942755543e-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": 0.006752799730747938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033896766020916403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016895275842398406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02151737492531538, + "weighted_rkd_loss": 0.008447637921199203 + }, + { + "epoch": 4.790619765494137, + "flow_matching_loss": 0.018972883094102146, + "grad_norm": 0.21053996682167053, + "learning_rate": 1.2687784650817616e-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": 0.006391771929338574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025758426752872767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015359386708587408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018972883094102146, + "weighted_rkd_loss": 0.007679693354293704 + }, + { + "epoch": 4.807370184254607, + "flow_matching_loss": 0.020155242085456847, + "grad_norm": 0.15815846621990204, + "learning_rate": 1.2589813502669603e-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": 0.0053243459668010475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020828464068472384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012731538340449332, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020155242085456847, + "weighted_rkd_loss": 0.006365769170224666 + }, + { + "epoch": 4.824120603015075, + "flow_matching_loss": 0.019101917929947376, + "grad_norm": 0.14777660369873047, + "learning_rate": 1.2491947783886907e-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": 0.006484691519290209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031254583853296937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016094841342419387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019101917929947376, + "weighted_rkd_loss": 0.008047420671209693 + }, + { + "epoch": 4.840871021775545, + "flow_matching_loss": 0.01791600426658988, + "grad_norm": 0.12246904522180557, + "learning_rate": 1.2394191775433106e-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": 0.005786749953404069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028565174899995325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014430017489939929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01791600426658988, + "weighted_rkd_loss": 0.007215008744969964 + }, + { + "epoch": 4.857621440536013, + "flow_matching_loss": 0.016301574185490608, + "grad_norm": 0.1843765676021576, + "learning_rate": 1.2296549753472696e-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": 0.005566553748212755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026679943315684796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013801102060824633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016301574185490608, + "weighted_rkd_loss": 0.006900551030412316 + }, + { + "epoch": 4.874371859296483, + "flow_matching_loss": 0.01598440548405051, + "grad_norm": 0.17817851901054382, + "learning_rate": 1.2199025989184032e-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": 0.00497903733048588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021776941372081637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012135768868029117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01598440548405051, + "weighted_rkd_loss": 0.006067884434014559 + }, + { + "epoch": 4.891122278056951, + "flow_matching_loss": 0.018774052895605565, + "grad_norm": 0.13188794255256653, + "learning_rate": 1.21016247485725e-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": 0.005346168461255729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025814786669798194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013273815717548132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018774052895605565, + "weighted_rkd_loss": 0.006636907858774066 + }, + { + "epoch": 4.907872696817421, + "flow_matching_loss": 0.018068937305361032, + "grad_norm": 0.1632096916437149, + "learning_rate": 1.2004350292283897e-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": 0.005146085587330163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023135139606893064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012605685275048018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018068937305361032, + "weighted_rkd_loss": 0.006302842637524009 + }, + { + "epoch": 4.924623115577889, + "flow_matching_loss": 0.017883090209215878, + "grad_norm": 0.15789315104484558, + "learning_rate": 1.1907206875418068e-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": 0.006365909287706017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027634695288725197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01549528781324625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017883090209215878, + "weighted_rkd_loss": 0.007747643906623125 + }, + { + "epoch": 4.941373534338359, + "flow_matching_loss": 0.018130379170179366, + "grad_norm": 0.18408702313899994, + "learning_rate": 1.1810198747342769e-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": 0.0056016219314187765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025636515463702382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01376689551398158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018130379170179366, + "weighted_rkd_loss": 0.00688344775699079 + }, + { + "epoch": 4.958123953098827, + "flow_matching_loss": 0.015793959517031907, + "grad_norm": 0.1873013824224472, + "learning_rate": 1.1713330151507786e-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": 0.006022406509146094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025855567771941425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014630369748920203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015793959517031907, + "weighted_rkd_loss": 0.007315184874460101 + }, + { + "epoch": 4.974874371859297, + "flow_matching_loss": 0.01843378497287631, + "grad_norm": 0.1450514942407608, + "learning_rate": 1.161660532525931e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063908693846315145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002764756348915398, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015546495281159877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01843378497287631, + "weighted_rkd_loss": 0.0077732476405799385 + }, + { + "epoch": 4.991624790619765, + "flow_matching_loss": 0.015466341562569142, + "grad_norm": 0.17179778218269348, + "learning_rate": 1.1520028499654584e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060694361571222545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024145430303178727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014553415589034557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015466341562569142, + "weighted_rkd_loss": 0.007276707794517279 + }, + { + "epoch": 5.008375209380235, + "flow_matching_loss": 0.017944001685827972, + "grad_norm": 0.21506550908088684, + "learning_rate": 1.1423603899276833e-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": 0.006832077959552407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002628091024234891, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016292246989905834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017944001685827972, + "weighted_rkd_loss": 0.008146123494952917 + }, + { + "epoch": 5.025125628140704, + "flow_matching_loss": 0.017549295723438264, + "grad_norm": 0.26095050573349, + "learning_rate": 1.1327335742050442e-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": 0.006050047976896167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002698522701393813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01479861866682768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017549295723438264, + "weighted_rkd_loss": 0.00739930933341384 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..27700cac5d292d4d42897b3c2e80bdc58e55c60e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5719731235192254, + -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-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a3758b2dac9cab1274647a0a3e4647bb197ec7fc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6f702e66e76c9c2a5773af5ec6b03a2aabb911ab03f04ce0b6ef1d2a8dca66a +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b23085757fd91fae834f4c6403f9a84dfd99ffc0 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6242297ba3d089711e642949f60242f06dbd7988bca7c9b6335962c0a8817a25 +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/checkpoint-5000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..db6ae6d46dd4ecc3c861a740372288147ed3401d --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_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": 0.20356450378894805, + "grad_norm": 1.0490323305130005, + "learning_rate": 1.08e-06, + "loss": 0.4004, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1398262470960617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11395216882228851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.39360466301441194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.20356450378894805, + "weighted_rkd_loss": 0.19680233150720597 + }, + { + "epoch": 0.03350083752093802, + "flow_matching_loss": 0.21135056614875794, + "grad_norm": 0.8279632925987244, + "learning_rate": 2.28e-06, + "loss": 0.3916, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12995845153927804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10064354985952377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3605604529380798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.21135056614875794, + "weighted_rkd_loss": 0.1802802264690399 + }, + { + "epoch": 0.05025125628140704, + "flow_matching_loss": 0.19959198087453842, + "grad_norm": 0.8481504321098328, + "learning_rate": 3.48e-06, + "loss": 0.3868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13302332013845444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1083116926252842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.374358332157135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.19959198087453842, + "weighted_rkd_loss": 0.1871791660785675 + }, + { + "epoch": 0.06700167504187604, + "flow_matching_loss": 0.21401427686214447, + "grad_norm": 1.1905933618545532, + "learning_rate": 4.68e-06, + "loss": 0.3892, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12562137991189956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09915168732404708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3503944456577301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.21401427686214447, + "weighted_rkd_loss": 0.17519722282886505 + }, + { + "epoch": 0.08375209380234507, + "flow_matching_loss": 0.18060285598039627, + "grad_norm": 0.9187169075012207, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11310266405344009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08473661690950393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3109419494867325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.18060285598039627, + "weighted_rkd_loss": 0.15547097474336624 + }, + { + "epoch": 0.10050251256281408, + "flow_matching_loss": 0.1746109426021576, + "grad_norm": 0.9370708465576172, + "learning_rate": 7.08e-06, + "loss": 0.3298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11362355500459671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08309690579771996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.31034401059150696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1746109426021576, + "weighted_rkd_loss": 0.15517200529575348 + }, + { + "epoch": 0.11725293132328309, + "flow_matching_loss": 0.16165669709444047, + "grad_norm": 0.7368708848953247, + "learning_rate": 8.28e-06, + "loss": 0.3148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11314839944243431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07991818860173225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30621498823165894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.16165669709444047, + "weighted_rkd_loss": 0.15310749411582947 + }, + { + "epoch": 0.13400335008375208, + "flow_matching_loss": 0.15373115316033364, + "grad_norm": 0.8529441356658936, + "learning_rate": 9.48e-06, + "loss": 0.2947, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10446386188268661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0729650903493166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.28189281225204466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.15373115316033364, + "weighted_rkd_loss": 0.14094640612602233 + }, + { + "epoch": 0.1507537688442211, + "flow_matching_loss": 0.12661252617836, + "grad_norm": 0.7936683893203735, + "learning_rate": 1.068e-05, + "loss": 0.2582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09823484048247337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06663991399109363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2631095930933952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.12661252617836, + "weighted_rkd_loss": 0.1315547965466976 + }, + { + "epoch": 0.16750418760469013, + "flow_matching_loss": 0.10844202637672425, + "grad_norm": 0.6985727548599243, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.2325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09290647432208061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.062390904128551486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24820385426282882, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.10844202637672425, + "weighted_rkd_loss": 0.12410192713141441 + }, + { + "epoch": 0.18425460636515914, + "flow_matching_loss": 0.09378867372870445, + "grad_norm": 0.6144883036613464, + "learning_rate": 1.308e-05, + "loss": 0.2126, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08944455981254577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05871000774204731, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23759912848472595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09378867372870445, + "weighted_rkd_loss": 0.11879956424236297 + }, + { + "epoch": 0.20100502512562815, + "flow_matching_loss": 0.08636006340384483, + "grad_norm": 0.519714891910553, + "learning_rate": 1.428e-05, + "loss": 0.1946, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08254799991846085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0513349212706089, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21643092334270478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08636006340384483, + "weighted_rkd_loss": 0.10821546167135239 + }, + { + "epoch": 0.21775544388609716, + "flow_matching_loss": 0.07800828218460083, + "grad_norm": 0.6437940001487732, + "learning_rate": 1.548e-05, + "loss": 0.1787, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07631000094115734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.048831694200634955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20145169496536255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07800828218460083, + "weighted_rkd_loss": 0.10072584748268128 + }, + { + "epoch": 0.23450586264656617, + "flow_matching_loss": 0.08341701701283455, + "grad_norm": 0.6835018992424011, + "learning_rate": 1.6680000000000003e-05, + "loss": 0.1816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0760519951581955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04422654993832111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19633053839206696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08341701701283455, + "weighted_rkd_loss": 0.09816526919603348 + }, + { + "epoch": 0.25125628140703515, + "flow_matching_loss": 0.07501983791589736, + "grad_norm": 0.6060196161270142, + "learning_rate": 1.7879999999999998e-05, + "loss": 0.1644, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06893355883657933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04092012867331505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17878724485635758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07501983791589736, + "weighted_rkd_loss": 0.08939362242817879 + }, + { + "epoch": 0.26800670016750416, + "flow_matching_loss": 0.07575965747237205, + "grad_norm": 0.5369148254394531, + "learning_rate": 1.908e-05, + "loss": 0.1611, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06494721546769142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.040794138610363004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17068857252597808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07575965747237205, + "weighted_rkd_loss": 0.08534428626298904 + }, + { + "epoch": 0.2847571189279732, + "flow_matching_loss": 0.07186435237526893, + "grad_norm": 0.4956815242767334, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06285238228738307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03799100033938885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1636957660317421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07186435237526893, + "weighted_rkd_loss": 0.08184788301587105 + }, + { + "epoch": 0.3015075376884422, + "flow_matching_loss": 0.06954732127487659, + "grad_norm": 0.4060089886188507, + "learning_rate": 2.148e-05, + "loss": 0.1457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05899851806461811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034271294809877874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1522683322429657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06954732127487659, + "weighted_rkd_loss": 0.07613416612148285 + }, + { + "epoch": 0.31825795644891125, + "flow_matching_loss": 0.07116162888705731, + "grad_norm": 0.4704577922821045, + "learning_rate": 2.268e-05, + "loss": 0.1464, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05825379975140095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033995861932635305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15050345957279204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07116162888705731, + "weighted_rkd_loss": 0.07525172978639602 + }, + { + "epoch": 0.33500837520938026, + "flow_matching_loss": 0.07152839414775372, + "grad_norm": 0.4596906900405884, + "learning_rate": 2.3880000000000002e-05, + "loss": 0.1431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05577990598976612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03162169083952904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14318150281906128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07152839414775372, + "weighted_rkd_loss": 0.07159075140953064 + }, + { + "epoch": 0.35175879396984927, + "flow_matching_loss": 0.06675980575382709, + "grad_norm": 0.46783247590065, + "learning_rate": 2.508e-05, + "loss": 0.1351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.052671773731708525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0314154077321291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13675895631313323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06675980575382709, + "weighted_rkd_loss": 0.06837947815656661 + }, + { + "epoch": 0.3685092127303183, + "flow_matching_loss": 0.06015884168446064, + "grad_norm": 0.4413231313228607, + "learning_rate": 2.628e-05, + "loss": 0.12, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04683331362903118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02598679680377245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11965342536568642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06015884168446064, + "weighted_rkd_loss": 0.05982671268284321 + }, + { + "epoch": 0.3852596314907873, + "flow_matching_loss": 0.05922078154981136, + "grad_norm": 0.4952040910720825, + "learning_rate": 2.748e-05, + "loss": 0.1187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04629960469901562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02644723802804947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1190464474260807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05922078154981136, + "weighted_rkd_loss": 0.05952322371304035 + }, + { + "epoch": 0.4020100502512563, + "flow_matching_loss": 0.06449385173618793, + "grad_norm": 0.48206594586372375, + "learning_rate": 2.868e-05, + "loss": 0.1224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04490597955882549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025913392566144465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11572535187005997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06449385173618793, + "weighted_rkd_loss": 0.057862675935029986 + }, + { + "epoch": 0.4187604690117253, + "flow_matching_loss": 0.054776575043797494, + "grad_norm": 0.40395382046699524, + "learning_rate": 2.9880000000000002e-05, + "loss": 0.1104, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04352367855608463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024162129126489164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11120948567986488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.054776575043797494, + "weighted_rkd_loss": 0.05560474283993244 + }, + { + "epoch": 0.4355108877721943, + "flow_matching_loss": 0.05850586108863354, + "grad_norm": 0.4056771993637085, + "learning_rate": 2.9999734259635676e-05, + "loss": 0.1118, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041912122443318364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022748018987476824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10657226219773293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05850586108863354, + "weighted_rkd_loss": 0.053286131098866464 + }, + { + "epoch": 0.45226130653266333, + "flow_matching_loss": 0.062202782928943635, + "grad_norm": 0.4282694160938263, + "learning_rate": 2.9998815663057245e-05, + "loss": 0.1095, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036747928336262704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02105743968859315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0945532962679863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.062202782928943635, + "weighted_rkd_loss": 0.04727664813399315 + }, + { + "epoch": 0.46901172529313234, + "flow_matching_loss": 0.051905983313918116, + "grad_norm": 0.39980289340019226, + "learning_rate": 2.999724096969228e-05, + "loss": 0.0978, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03628676719963551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019146752916276454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09172028750181198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.051905983313918116, + "weighted_rkd_loss": 0.04586014375090599 + }, + { + "epoch": 0.48576214405360135, + "flow_matching_loss": 0.05202174261212349, + "grad_norm": 0.42642244696617126, + "learning_rate": 2.9995010248422984e-05, + "loss": 0.0982, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03656212091445923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01933057326823473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0924548164010048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05202174261212349, + "weighted_rkd_loss": 0.0462274082005024 + }, + { + "epoch": 0.5025125628140703, + "flow_matching_loss": 0.048975753411650655, + "grad_norm": 0.3835107982158661, + "learning_rate": 2.999212359682832e-05, + "loss": 0.0927, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034414700232446194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018641455005854367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08747085481882096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.048975753411650655, + "weighted_rkd_loss": 0.04373542740941048 + }, + { + "epoch": 0.5192629815745393, + "flow_matching_loss": 0.05028046742081642, + "grad_norm": 0.3581988513469696, + "learning_rate": 2.9988581141179787e-05, + "loss": 0.0919, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03313282225281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016972948610782624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08323859423398972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05028046742081642, + "weighted_rkd_loss": 0.04161929711699486 + }, + { + "epoch": 0.5360134003350083, + "flow_matching_loss": 0.048362468183040616, + "grad_norm": 0.3743322193622589, + "learning_rate": 2.998438303643587e-05, + "loss": 0.0896, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03231747522950172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017821643501520157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0824565939605236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.048362468183040616, + "weighted_rkd_loss": 0.0412282969802618 + }, + { + "epoch": 0.5527638190954773, + "flow_matching_loss": 0.04962967894971371, + "grad_norm": 0.37111181020736694, + "learning_rate": 2.997952946623527e-05, + "loss": 0.0898, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031945614889264104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016547599248588084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08043882995843887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04962967894971371, + "weighted_rkd_loss": 0.04021941497921944 + }, + { + "epoch": 0.5695142378559463, + "flow_matching_loss": 0.050852949172258376, + "grad_norm": 0.418589323759079, + "learning_rate": 2.9974020642888875e-05, + "loss": 0.0892, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03034352362155914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016099225915968417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07678627483546734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050852949172258376, + "weighted_rkd_loss": 0.03839313741773367 + }, + { + "epoch": 0.5862646566164154, + "flow_matching_loss": 0.05294822007417679, + "grad_norm": 0.4339646100997925, + "learning_rate": 2.9967856807370458e-05, + "loss": 0.0964, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03385459519922733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019144115131348373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08685330525040627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05294822007417679, + "weighted_rkd_loss": 0.043426652625203135 + }, + { + "epoch": 0.6030150753768844, + "flow_matching_loss": 0.04927732534706593, + "grad_norm": 0.39898374676704407, + "learning_rate": 2.996103822930615e-05, + "loss": 0.09, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03226359337568283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01696971897035837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08149690553545952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04927732534706593, + "weighted_rkd_loss": 0.04074845276772976 + }, + { + "epoch": 0.6197654941373534, + "flow_matching_loss": 0.04757112897932529, + "grad_norm": 0.3817186653614044, + "learning_rate": 2.9953565206962653e-05, + "loss": 0.0851, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029814749024808407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015488553512841462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07511805146932601, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04757112897932529, + "weighted_rkd_loss": 0.03755902573466301 + }, + { + "epoch": 0.6365159128978225, + "flow_matching_loss": 0.04780330918729305, + "grad_norm": 0.35171931982040405, + "learning_rate": 2.9945438067234172e-05, + "loss": 0.0839, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028515780717134474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01506363358348608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07209519408643246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04780330918729305, + "weighted_rkd_loss": 0.03604759704321623 + }, + { + "epoch": 0.6532663316582915, + "flow_matching_loss": 0.04430910907685757, + "grad_norm": 0.39219847321510315, + "learning_rate": 2.9936657165628124e-05, + "loss": 0.0793, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027778214029967784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01443298477679491, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06998941153287888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04430910907685757, + "weighted_rkd_loss": 0.03499470576643944 + }, + { + "epoch": 0.6700167504187605, + "flow_matching_loss": 0.04167022332549095, + "grad_norm": 0.3763214945793152, + "learning_rate": 2.992722288624959e-05, + "loss": 0.0763, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02743688691407442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014438783749938011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06931255795061589, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04167022332549095, + "weighted_rkd_loss": 0.034656278975307944 + }, + { + "epoch": 0.6867671691792295, + "flow_matching_loss": 0.04623462222516537, + "grad_norm": 0.3273172974586487, + "learning_rate": 2.9917135641784515e-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": 0.026866826787590982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0142863011918962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06801995486021042, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04623462222516537, + "weighted_rkd_loss": 0.03400997743010521 + }, + { + "epoch": 0.7035175879396985, + "flow_matching_loss": 0.042290421202778816, + "grad_norm": 0.3280941843986511, + "learning_rate": 2.9906395873481655e-05, + "loss": 0.0747, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025736703351140023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013280625827610493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06475403234362602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042290421202778816, + "weighted_rkd_loss": 0.03237701617181301 + }, + { + "epoch": 0.7202680067001676, + "flow_matching_loss": 0.04566530957818031, + "grad_norm": 0.3375310003757477, + "learning_rate": 2.989500405113326e-05, + "loss": 0.0792, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02661526370793581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013754750974476338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06698527745902538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04566530957818031, + "weighted_rkd_loss": 0.03349263872951269 + }, + { + "epoch": 0.7370184254606366, + "flow_matching_loss": 0.041253012418746945, + "grad_norm": 0.36417585611343384, + "learning_rate": 2.9882960673054547e-05, + "loss": 0.0736, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025740651600062846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013157044537365437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06463834792375564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.041253012418746945, + "weighted_rkd_loss": 0.03231917396187782 + }, + { + "epoch": 0.7537688442211056, + "flow_matching_loss": 0.0429521806538105, + "grad_norm": 0.31173717975616455, + "learning_rate": 2.9870266266061886e-05, + "loss": 0.0734, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02408239897340536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012752699572592974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060917497426271436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0429521806538105, + "weighted_rkd_loss": 0.030458748713135718 + }, + { + "epoch": 0.7705192629815746, + "flow_matching_loss": 0.04328401610255241, + "grad_norm": 0.32832983136177063, + "learning_rate": 2.9856921385449766e-05, + "loss": 0.0763, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02588608395308256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01416186960414052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06593403704464436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04328401610255241, + "weighted_rkd_loss": 0.03296701852232218 + }, + { + "epoch": 0.7872696817420436, + "flow_matching_loss": 0.042371243983507154, + "grad_norm": 0.322933167219162, + "learning_rate": 2.9842926614966502e-05, + "loss": 0.0717, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023254065960645675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012198315933346748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0587064478546381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042371243983507154, + "weighted_rkd_loss": 0.02935322392731905 + }, + { + "epoch": 0.8040201005025126, + "flow_matching_loss": 0.036948934011161325, + "grad_norm": 0.37090227007865906, + "learning_rate": 2.982828256678868e-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": 0.023912062868475914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012023937422782183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05984806343913078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.036948934011161325, + "weighted_rkd_loss": 0.02992403171956539 + }, + { + "epoch": 0.8207705192629816, + "flow_matching_loss": 0.04164762869477272, + "grad_norm": 0.2612155079841614, + "learning_rate": 2.9812989881494412e-05, + "loss": 0.0693, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022154004499316215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011067771259695292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05537578016519547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04164762869477272, + "weighted_rkd_loss": 0.027687890082597734 + }, + { + "epoch": 0.8375209380234506, + "flow_matching_loss": 0.041829455830156806, + "grad_norm": 0.3849196434020996, + "learning_rate": 2.9797049228035296e-05, + "loss": 0.0696, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022308938018977643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010920075606554747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05553795211017132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.041829455830156806, + "weighted_rkd_loss": 0.02776897605508566 + }, + { + "epoch": 0.8542713567839196, + "flow_matching_loss": 0.039186670817434786, + "grad_norm": 0.2965879738330841, + "learning_rate": 2.978046130370715e-05, + "loss": 0.0642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02009088359773159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009756325092166663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04993809200823307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.039186670817434786, + "weighted_rkd_loss": 0.024969046004116535 + }, + { + "epoch": 0.8710217755443886, + "flow_matching_loss": 0.03997811581939459, + "grad_norm": 0.3125465512275696, + "learning_rate": 2.9763226834119535e-05, + "loss": 0.0675, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0217224583029747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011559972260147332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05500488877296448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03997811581939459, + "weighted_rkd_loss": 0.02750244438648224 + }, + { + "epoch": 0.8877721943048577, + "flow_matching_loss": 0.037035029381513596, + "grad_norm": 0.28196239471435547, + "learning_rate": 2.9745346573163963e-05, + "loss": 0.0683, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024690970405936243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013126031681895255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06250797249376774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037035029381513596, + "weighted_rkd_loss": 0.03125398624688387 + }, + { + "epoch": 0.9045226130653267, + "flow_matching_loss": 0.03601754531264305, + "grad_norm": 0.3579603135585785, + "learning_rate": 2.9726821302980987e-05, + "loss": 0.0653, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023435172624886036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011736051738262176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05860639661550522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03601754531264305, + "weighted_rkd_loss": 0.02930319830775261 + }, + { + "epoch": 0.9212730318257957, + "flow_matching_loss": 0.035668034479022025, + "grad_norm": 0.33571353554725647, + "learning_rate": 2.9707651833925935e-05, + "loss": 0.0603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01971091693267226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009920926159247756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049342759884893894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035668034479022025, + "weighted_rkd_loss": 0.024671379942446947 + }, + { + "epoch": 0.9380234505862647, + "flow_matching_loss": 0.04034342188388109, + "grad_norm": 0.37223154306411743, + "learning_rate": 2.968783900453348e-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": 0.02227401491254568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010327170602977276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054875200241804124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04034342188388109, + "weighted_rkd_loss": 0.027437600120902062 + }, + { + "epoch": 0.9547738693467337, + "flow_matching_loss": 0.03982900306582451, + "grad_norm": 0.3271760940551758, + "learning_rate": 2.9667383681480968e-05, + "loss": 0.0647, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019855904299765824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010042787250131368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04975459612905979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03982900306582451, + "weighted_rkd_loss": 0.024877298064529894 + }, + { + "epoch": 0.9715242881072027, + "flow_matching_loss": 0.04037307444959879, + "grad_norm": 0.3598499596118927, + "learning_rate": 2.9646286759550493e-05, + "loss": 0.0652, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01992246611043811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009732879279181362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04957781210541725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04037307444959879, + "weighted_rkd_loss": 0.024788906052708625 + }, + { + "epoch": 0.9882747068676717, + "flow_matching_loss": 0.03921057693660259, + "grad_norm": 0.4509306848049164, + "learning_rate": 2.9624549161589764e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018727950938045978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009081155760213733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04653705805540085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03921057693660259, + "weighted_rkd_loss": 0.023268529027700425 + }, + { + "epoch": 1.0050251256281406, + "flow_matching_loss": 0.03464273381978274, + "grad_norm": 0.3406599164009094, + "learning_rate": 2.9602171838471745e-05, + "loss": 0.0614, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02158996183425188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01035784943960607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05353777259588242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03464273381978274, + "weighted_rkd_loss": 0.02676888629794121 + }, + { + "epoch": 1.0217755443886096, + "flow_matching_loss": 0.032238774560391904, + "grad_norm": 0.3566734194755554, + "learning_rate": 2.957915576905303e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019271148182451726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010775866499170662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049318162724375725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032238774560391904, + "weighted_rkd_loss": 0.024659081362187862 + }, + { + "epoch": 1.0385259631490786, + "flow_matching_loss": 0.033852733485400674, + "grad_norm": 0.3310783803462982, + "learning_rate": 2.9555501960131058e-05, + "loss": 0.0578, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019164290837943555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009481828706339002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047810410335659984, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033852733485400674, + "weighted_rkd_loss": 0.023905205167829992 + }, + { + "epoch": 1.0552763819095476, + "flow_matching_loss": 0.03569246102124453, + "grad_norm": 0.29758018255233765, + "learning_rate": 2.9531211446400068e-05, + "loss": 0.0619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020986123383045195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010503490781411528, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052475737407803535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03569246102124453, + "weighted_rkd_loss": 0.026237868703901768 + }, + { + "epoch": 1.0720268006700167, + "flow_matching_loss": 0.03429760187864304, + "grad_norm": 0.34033986926078796, + "learning_rate": 2.9506285290405813e-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": 0.019663255382329225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009310608357191085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048637119308114055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03429760187864304, + "weighted_rkd_loss": 0.024318559654057027 + }, + { + "epoch": 1.0887772194304857, + "flow_matching_loss": 0.03442421518266201, + "grad_norm": 0.3421791195869446, + "learning_rate": 2.948072458249911e-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": 0.018693058285862207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00901120575144887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04639732241630554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03442421518266201, + "weighted_rkd_loss": 0.02319866120815277 + }, + { + "epoch": 1.1055276381909547, + "flow_matching_loss": 0.033443206921219826, + "grad_norm": 0.3156006932258606, + "learning_rate": 2.9454530440788125e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020682687778025864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009997753286734223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05136312879621983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033443206921219826, + "weighted_rkd_loss": 0.025681564398109914 + }, + { + "epoch": 1.1222780569514237, + "flow_matching_loss": 0.035716149397194384, + "grad_norm": 0.3609527349472046, + "learning_rate": 2.9427704011089474e-05, + "loss": 0.0601, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01966800969094038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009427881613373757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04876390062272549, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035716149397194384, + "weighted_rkd_loss": 0.024381950311362744 + }, + { + "epoch": 1.1390284757118927, + "flow_matching_loss": 0.03611791767179966, + "grad_norm": 0.29612571001052856, + "learning_rate": 2.94002464668781e-05, + "loss": 0.0577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017426843661814927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008403770998120308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04325745925307274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03611791767179966, + "weighted_rkd_loss": 0.02162872962653637 + }, + { + "epoch": 1.1557788944723617, + "flow_matching_loss": 0.03446404431015253, + "grad_norm": 0.26498663425445557, + "learning_rate": 2.9372159009235936e-05, + "loss": 0.0559, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01693731788545847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008955044439062476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042829679884016515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03446404431015253, + "weighted_rkd_loss": 0.021414839942008258 + }, + { + "epoch": 1.1725293132328307, + "flow_matching_loss": 0.03775186575949192, + "grad_norm": 0.2350740283727646, + "learning_rate": 2.9343442866799365e-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": 0.01552609372884035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007015585526823997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03806777279824018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03775186575949192, + "weighted_rkd_loss": 0.01903388639912009 + }, + { + "epoch": 1.1892797319932997, + "flow_matching_loss": 0.034802002087235454, + "grad_norm": 0.27234312891960144, + "learning_rate": 2.9314099295705495e-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": 0.01639019064605236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008023013034835458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04080339428037405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034802002087235454, + "weighted_rkd_loss": 0.020401697140187025 + }, + { + "epoch": 1.2060301507537687, + "flow_matching_loss": 0.033095156960189345, + "grad_norm": 0.2403380125761032, + "learning_rate": 2.928412957953718e-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": 0.016718606650829314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00849555665627122, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04193276967853308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033095156960189345, + "weighted_rkd_loss": 0.02096638483926654 + }, + { + "epoch": 1.2227805695142377, + "flow_matching_loss": 0.03338757548481226, + "grad_norm": 0.34521374106407166, + "learning_rate": 2.925353502926689e-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": 0.016283851675689222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007864421885460614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04043212570250034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03338757548481226, + "weighted_rkd_loss": 0.02021606285125017 + }, + { + "epoch": 1.2395309882747068, + "flow_matching_loss": 0.029536503180861474, + "grad_norm": 0.313785195350647, + "learning_rate": 2.922231698319937e-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": 0.015204551909118891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007029514294117689, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03743861857801676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029536503180861474, + "weighted_rkd_loss": 0.01871930928900838 + }, + { + "epoch": 1.2562814070351758, + "flow_matching_loss": 0.031847203895449636, + "grad_norm": 0.31126895546913147, + "learning_rate": 2.9190476806913088e-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": 0.015663769003003834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007163937482982874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038491476140916345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031847203895449636, + "weighted_rkd_loss": 0.019245738070458172 + }, + { + "epoch": 1.2730318257956448, + "flow_matching_loss": 0.03111131377518177, + "grad_norm": 0.24699684977531433, + "learning_rate": 2.9158015893200504e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015276441443711519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007583758374676108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038136640936136244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03111131377518177, + "weighted_rkd_loss": 0.019068320468068122 + }, + { + "epoch": 1.2897822445561138, + "flow_matching_loss": 0.03199950084090233, + "grad_norm": 0.26863768696784973, + "learning_rate": 2.9124935662007137e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014959248062223197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007236627629026771, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03715512380003929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03199950084090233, + "weighted_rkd_loss": 0.018577561900019645 + }, + { + "epoch": 1.3065326633165828, + "flow_matching_loss": 0.031591197848320006, + "grad_norm": 0.27149420976638794, + "learning_rate": 2.9091237560369467e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01730004707351327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007396387634798884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041996481455862524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031591197848320006, + "weighted_rkd_loss": 0.020998240727931262 + }, + { + "epoch": 1.3232830820770518, + "flow_matching_loss": 0.03199263922870159, + "grad_norm": 0.23743563890457153, + "learning_rate": 2.905692306235163e-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": 0.014362255297601222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007487459061667323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03621196895837784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03199263922870159, + "weighted_rkd_loss": 0.01810598447918892 + }, + { + "epoch": 1.3400335008375208, + "flow_matching_loss": 0.032203089259564874, + "grad_norm": 0.26545020937919617, + "learning_rate": 2.9021993668980925e-05, + "loss": 0.0509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015238793008029461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006904208846390247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03738179467618465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032203089259564874, + "weighted_rkd_loss": 0.018690897338092326 + }, + { + "epoch": 1.3567839195979898, + "flow_matching_loss": 0.03435746598988772, + "grad_norm": 0.3084009289741516, + "learning_rate": 2.8986450908182185e-05, + "loss": 0.0519, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014165364857763051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00666372743435204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03499445673078298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03435746598988772, + "weighted_rkd_loss": 0.01749722836539149 + }, + { + "epoch": 1.3735343383584588, + "flow_matching_loss": 0.03177093621343374, + "grad_norm": 0.26156696677207947, + "learning_rate": 2.8950296334710902e-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": 0.01559155946597457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0073688535019755365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03855197243392468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03177093621343374, + "weighted_rkd_loss": 0.01927598621696234 + }, + { + "epoch": 1.3902847571189278, + "flow_matching_loss": 0.03415762279182673, + "grad_norm": 0.3244900107383728, + "learning_rate": 2.8913531530085248e-05, + "loss": 0.054, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015872035082429646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00787778990343213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03962186016142368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03415762279182673, + "weighted_rkd_loss": 0.01981093008071184 + }, + { + "epoch": 1.4070351758793969, + "flow_matching_loss": 0.031543026119470595, + "grad_norm": 0.24670489132404327, + "learning_rate": 2.887615810251687e-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": 0.014911935105919838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007098326459527016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03692219667136669, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031543026119470595, + "weighted_rkd_loss": 0.018461098335683344 + }, + { + "epoch": 1.4237855946398659, + "flow_matching_loss": 0.030299522541463376, + "grad_norm": 0.29750338196754456, + "learning_rate": 2.8838177686840577e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013420206494629383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006395753938704729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033236166834831236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030299522541463376, + "weighted_rkd_loss": 0.016618083417415618 + }, + { + "epoch": 1.4405360134003349, + "flow_matching_loss": 0.03428787123411894, + "grad_norm": 0.2844262421131134, + "learning_rate": 2.8799591944442776e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014772494789212942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007544750696979463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0370897401124239, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03428787123411894, + "weighted_rkd_loss": 0.01854487005621195 + }, + { + "epoch": 1.457286432160804, + "flow_matching_loss": 0.028464125469326973, + "grad_norm": 0.2586044669151306, + "learning_rate": 2.8760402563188848e-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": 0.015234122145920992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007229142403230071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03769738636910915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028464125469326973, + "weighted_rkd_loss": 0.018848693184554576 + }, + { + "epoch": 1.474036850921273, + "flow_matching_loss": 0.03240107372403145, + "grad_norm": 0.3451785743236542, + "learning_rate": 2.8720611257349285e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01441895104944706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007548008835874498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036385910399258134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03240107372403145, + "weighted_rkd_loss": 0.018192955199629067 + }, + { + "epoch": 1.490787269681742, + "flow_matching_loss": 0.031261354126036166, + "grad_norm": 0.2677576243877411, + "learning_rate": 2.868021976752471e-05, + "loss": 0.0504, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015447228122502566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007448036875575781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03834249284118414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031261354126036166, + "weighted_rkd_loss": 0.01917124642059207 + }, + { + "epoch": 1.507537688442211, + "flow_matching_loss": 0.029524659179151057, + "grad_norm": 0.2589884400367737, + "learning_rate": 2.8639229860569738e-05, + "loss": 0.0467, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013584412820637225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007130756136029958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03429958187043667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029524659179151057, + "weighted_rkd_loss": 0.017149790935218335 + }, + { + "epoch": 1.52428810720268, + "flow_matching_loss": 0.030357540026307105, + "grad_norm": 0.2322537899017334, + "learning_rate": 2.8597643329515688e-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": 0.015393463242799044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007288130559027195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03807505648583174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030357540026307105, + "weighted_rkd_loss": 0.01903752824291587 + }, + { + "epoch": 1.541038525963149, + "flow_matching_loss": 0.033645805157721045, + "grad_norm": 0.2303711622953415, + "learning_rate": 2.855546199349214e-05, + "loss": 0.0525, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0153488309122622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070605908520519735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037758252769708636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033645805157721045, + "weighted_rkd_loss": 0.018879126384854318 + }, + { + "epoch": 1.557788944723618, + "flow_matching_loss": 0.029417644068598748, + "grad_norm": 0.33168652653694153, + "learning_rate": 2.85126876976474e-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": 0.013482074066996575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007291349256411195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03425549734383822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029417644068598748, + "weighted_rkd_loss": 0.01712774867191911 + }, + { + "epoch": 1.574539363484087, + "flow_matching_loss": 0.02936752066016197, + "grad_norm": 0.32166430354118347, + "learning_rate": 2.8469322313067726e-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": 0.014363228995352983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006481660483404994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035208118706941606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02936752066016197, + "weighted_rkd_loss": 0.017604059353470803 + }, + { + "epoch": 1.591289782244556, + "flow_matching_loss": 0.030744343996047974, + "grad_norm": 0.2652220129966736, + "learning_rate": 2.8425367736695522e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012823993526399135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006567410938441753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03221539761871099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030744343996047974, + "weighted_rkd_loss": 0.016107698809355496 + }, + { + "epoch": 1.608040201005025, + "flow_matching_loss": 0.027458759397268294, + "grad_norm": 0.2823118269443512, + "learning_rate": 2.838082589124636e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014246038906276227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006480111461132765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03497218918055296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027458759397268294, + "weighted_rkd_loss": 0.01748609459027648 + }, + { + "epoch": 1.624790619765494, + "flow_matching_loss": 0.02682869080454111, + "grad_norm": 0.27871882915496826, + "learning_rate": 2.8335698725124843e-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": 0.015229499991983175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0072166129481047395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037675613351166246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02682869080454111, + "weighted_rkd_loss": 0.018837806675583123 + }, + { + "epoch": 1.641541038525963, + "flow_matching_loss": 0.027610163576900958, + "grad_norm": 0.27080997824668884, + "learning_rate": 2.8289988212339422e-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": 0.014654815848916768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007433690736070276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03674332220107317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027610163576900958, + "weighted_rkd_loss": 0.018371661100536583 + }, + { + "epoch": 1.658291457286432, + "flow_matching_loss": 0.026477544382214547, + "grad_norm": 0.29053348302841187, + "learning_rate": 2.8243696352415997e-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": 0.014315103739500045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006944430945441127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035574638098478314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026477544382214547, + "weighted_rkd_loss": 0.017787319049239157 + }, + { + "epoch": 1.675041876046901, + "flow_matching_loss": 0.03149757422506809, + "grad_norm": 0.24316534399986267, + "learning_rate": 2.8196825170310495e-05, + "loss": 0.0489, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013985770475119352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006879748497158289, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034851289354264736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03149757422506809, + "weighted_rkd_loss": 0.017425644677132368 + }, + { + "epoch": 1.69179229480737, + "flow_matching_loss": 0.028719986975193023, + "grad_norm": 0.2793906033039093, + "learning_rate": 2.8149376716320256e-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": 0.014713972900062799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007857491169124842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03728543724864721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028719986975193023, + "weighted_rkd_loss": 0.018642718624323606 + }, + { + "epoch": 1.708542713567839, + "flow_matching_loss": 0.02652828861027956, + "grad_norm": 0.24760262668132782, + "learning_rate": 2.8101353065994373e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015948845073580742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008608205150812865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04050589501857758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02652828861027956, + "weighted_rkd_loss": 0.02025294750928879 + }, + { + "epoch": 1.725293132328308, + "flow_matching_loss": 0.02748129330575466, + "grad_norm": 0.2966518998146057, + "learning_rate": 2.8052756320042885e-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": 0.013009252492338419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00621428107842803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03223278634250164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02748129330575466, + "weighted_rkd_loss": 0.01611639317125082 + }, + { + "epoch": 1.742043551088777, + "flow_matching_loss": 0.02748612593859434, + "grad_norm": 0.2400917261838913, + "learning_rate": 2.8003588604244893e-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": 0.012675477657467127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006017118878662586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031368074007332325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02748612593859434, + "weighted_rkd_loss": 0.015684037003666162 + }, + { + "epoch": 1.758793969849246, + "flow_matching_loss": 0.0283775020390749, + "grad_norm": 0.23072923719882965, + "learning_rate": 2.7953852069355565e-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": 0.012886785715818406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007881310116499663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03365488145500421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0283775020390749, + "weighted_rkd_loss": 0.016827440727502106 + }, + { + "epoch": 1.775544388609715, + "flow_matching_loss": 0.028830614313483237, + "grad_norm": 0.19938981533050537, + "learning_rate": 2.790354889101206e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011350309383124113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005617117416113615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028317735902965067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028830614313483237, + "weighted_rkd_loss": 0.014158867951482534 + }, + { + "epoch": 1.792294807370184, + "flow_matching_loss": 0.02743749525398016, + "grad_norm": 0.20770491659641266, + "learning_rate": 2.7852681269638366e-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": 0.013101563602685929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006368691730313003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03257181905210018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02743749525398016, + "weighted_rkd_loss": 0.01628590952605009 + }, + { + "epoch": 1.809045226130653, + "flow_matching_loss": 0.025509939529001713, + "grad_norm": 0.23055991530418396, + "learning_rate": 2.780125143034902e-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": 0.01217031548731029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005663800402544439, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030004431121051313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025509939529001713, + "weighted_rkd_loss": 0.015002215560525656 + }, + { + "epoch": 1.8257956448911221, + "flow_matching_loss": 0.027603413537144662, + "grad_norm": 0.24784690141677856, + "learning_rate": 2.7749261622851812e-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": 0.012131767626851798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005731426505371928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02999496180564165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027603413537144662, + "weighted_rkd_loss": 0.014997480902820826 + }, + { + "epoch": 1.8425460636515911, + "flow_matching_loss": 0.02833728678524494, + "grad_norm": 0.2339480221271515, + "learning_rate": 2.7696714121349335e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011987958382815123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005508459242992103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029484376311302185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02833728678524494, + "weighted_rkd_loss": 0.014742188155651093 + }, + { + "epoch": 1.8592964824120601, + "flow_matching_loss": 0.028176490403711795, + "grad_norm": 0.21943725645542145, + "learning_rate": 2.764361122443954e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012409442663192749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006674071168527007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03149295635521412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028176490403711795, + "weighted_rkd_loss": 0.01574647817760706 + }, + { + "epoch": 1.8760469011725294, + "flow_matching_loss": 0.02386368773877621, + "grad_norm": 0.28784745931625366, + "learning_rate": 2.758995525501517e-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": 0.012525459285825491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006127721210941672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031178639829158784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02386368773877621, + "weighted_rkd_loss": 0.015589319914579392 + }, + { + "epoch": 1.8927973199329984, + "flow_matching_loss": 0.026401301100850105, + "grad_norm": 0.302274614572525, + "learning_rate": 2.753574856016215e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012502617854624986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005733438022434712, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03073867317289114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026401301100850105, + "weighted_rkd_loss": 0.01536933658644557 + }, + { + "epoch": 1.9095477386934674, + "flow_matching_loss": 0.025203173607587816, + "grad_norm": 0.22959397733211517, + "learning_rate": 2.7480993511056913e-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": 0.012079661898314952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005447410838678479, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029606734588742255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025203173607587816, + "weighted_rkd_loss": 0.014803367294371128 + }, + { + "epoch": 1.9262981574539364, + "flow_matching_loss": 0.02645403500646353, + "grad_norm": 0.20805855095386505, + "learning_rate": 2.7425692502862684e-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": 0.012591348122805357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005284116393886507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030466812662780285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02645403500646353, + "weighted_rkd_loss": 0.015233406331390142 + }, + { + "epoch": 1.9430485762144054, + "flow_matching_loss": 0.028687219321727752, + "grad_norm": 0.23209843039512634, + "learning_rate": 2.7369847954624728e-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": 0.01346703628078103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005913856625556946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032847929000854495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028687219321727752, + "weighted_rkd_loss": 0.016423964500427247 + }, + { + "epoch": 1.9597989949748744, + "flow_matching_loss": 0.0267184529453516, + "grad_norm": 0.20264476537704468, + "learning_rate": 2.731346230916448e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012482394743710757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005907515040598809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030872304178774356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0267184529453516, + "weighted_rkd_loss": 0.015436152089387178 + }, + { + "epoch": 1.9765494137353434, + "flow_matching_loss": 0.02526174336671829, + "grad_norm": 0.2656524181365967, + "learning_rate": 2.7256538032972746e-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": 0.011073235142976045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004903874779120088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02705034501850605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02526174336671829, + "weighted_rkd_loss": 0.013525172509253025 + }, + { + "epoch": 1.9932998324958124, + "flow_matching_loss": 0.024283120594918727, + "grad_norm": 0.18750469386577606, + "learning_rate": 2.7199077616101773e-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": 0.010565282870084047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005318943597376347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026449509151279926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024283120594918727, + "weighted_rkd_loss": 0.013224754575639963 + }, + { + "epoch": 2.0100502512562812, + "flow_matching_loss": 0.023823299817740916, + "grad_norm": 0.28387999534606934, + "learning_rate": 2.714108357205634e-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": 0.01328891459852457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061988724861294035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032776701636612414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023823299817740916, + "weighted_rkd_loss": 0.016388350818306207 + }, + { + "epoch": 2.0268006700167502, + "flow_matching_loss": 0.027536653354763983, + "grad_norm": 0.23609916865825653, + "learning_rate": 2.7082558437683803e-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": 0.010250479681417346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046427268767729405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025143686309456827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027536653354763983, + "weighted_rkd_loss": 0.012571843154728413 + }, + { + "epoch": 2.0435510887772192, + "flow_matching_loss": 0.025359765440225602, + "grad_norm": 0.2570204436779022, + "learning_rate": 2.702350477306315e-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": 0.012231682194396853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005420093261636794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02988345790654421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025359765440225602, + "weighted_rkd_loss": 0.014941728953272105 + }, + { + "epoch": 2.0603015075376883, + "flow_matching_loss": 0.02634656634181738, + "grad_norm": 0.22433871030807495, + "learning_rate": 2.6963925161392963e-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": 0.010704432427883149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047349676489830015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026143832691013814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02634656634181738, + "weighted_rkd_loss": 0.013071916345506907 + }, + { + "epoch": 2.0770519262981573, + "flow_matching_loss": 0.02522501554340124, + "grad_norm": 0.2080315798521042, + "learning_rate": 2.6903822208878476e-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": 0.010163325816392899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004687550477683544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02501420211046934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02522501554340124, + "weighted_rkd_loss": 0.01250710105523467 + }, + { + "epoch": 2.0938023450586263, + "flow_matching_loss": 0.02463902086019516, + "grad_norm": 0.25818106532096863, + "learning_rate": 2.6843198544617535e-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": 0.012039071694016457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006278580939397216, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030356723815202713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02463902086019516, + "weighted_rkd_loss": 0.015178361907601357 + }, + { + "epoch": 2.1105527638190953, + "flow_matching_loss": 0.027839330770075322, + "grad_norm": 0.2640082538127899, + "learning_rate": 2.6782056820485612e-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": 0.012098926212638616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005462716426700353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029660568572580814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027839330770075322, + "weighted_rkd_loss": 0.014830284286290407 + }, + { + "epoch": 2.1273031825795643, + "flow_matching_loss": 0.028276474215090275, + "grad_norm": 0.295583575963974, + "learning_rate": 2.672039971101979e-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": 0.013212186470627784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006356807774864137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032781181018799545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028276474215090275, + "weighted_rkd_loss": 0.016390590509399772 + }, + { + "epoch": 2.1440536013400333, + "flow_matching_loss": 0.026197852566838264, + "grad_norm": 0.20729024708271027, + "learning_rate": 2.665822991330178e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01121538388542831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005526613956317305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027957382053136824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026197852566838264, + "weighted_rkd_loss": 0.013978691026568412 + }, + { + "epoch": 2.1608040201005023, + "flow_matching_loss": 0.02445336002856493, + "grad_norm": 0.19470828771591187, + "learning_rate": 2.6595550146839934e-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": 0.011698291730135679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005225977627560496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02862256094813347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02445336002856493, + "weighted_rkd_loss": 0.014311280474066735 + }, + { + "epoch": 2.1775544388609713, + "flow_matching_loss": 0.02761701140552759, + "grad_norm": 0.29102823138237, + "learning_rate": 2.6532363153450287e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012762863840907811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00593646764755249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031462194956839086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02761701140552759, + "weighted_rkd_loss": 0.015731097478419543 + }, + { + "epoch": 2.1943048576214403, + "flow_matching_loss": 0.023142579942941666, + "grad_norm": 0.2832508087158203, + "learning_rate": 2.6468671697136634e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01395815759897232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007116776378825307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03503309153020382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023142579942941666, + "weighted_rkd_loss": 0.01751654576510191 + }, + { + "epoch": 2.2110552763819094, + "flow_matching_loss": 0.024112015962600708, + "grad_norm": 0.22076784074306488, + "learning_rate": 2.640447856396961e-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": 0.010843491833657026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005128723871894181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026815707236528395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024112015962600708, + "weighted_rkd_loss": 0.013407853618264198 + }, + { + "epoch": 2.2278056951423784, + "flow_matching_loss": 0.02480712588876486, + "grad_norm": 0.24121958017349243, + "learning_rate": 2.6339786561964805e-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": 0.010331244813278318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004812800046056509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025475289672613144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02480712588876486, + "weighted_rkd_loss": 0.012737644836306572 + }, + { + "epoch": 2.2445561139028474, + "flow_matching_loss": 0.02151898145675659, + "grad_norm": 0.2851496636867523, + "learning_rate": 2.6274598520959966e-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": 0.011607506917789579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005171733908355236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028386748023331167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02151898145675659, + "weighted_rkd_loss": 0.014193374011665583 + }, + { + "epoch": 2.2613065326633164, + "flow_matching_loss": 0.02450380623340607, + "grad_norm": 0.19669336080551147, + "learning_rate": 2.6208917292491185e-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": 0.011534726107493043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005064515001140535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02813396751880646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02450380623340607, + "weighted_rkd_loss": 0.01406698375940323 + }, + { + "epoch": 2.2780569514237854, + "flow_matching_loss": 0.0243028256110847, + "grad_norm": 0.22813156247138977, + "learning_rate": 2.6142745749668176e-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": 0.008905993402004242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004185553826391697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021997540444135665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0243028256110847, + "weighted_rkd_loss": 0.010998770222067833 + }, + { + "epoch": 2.2948073701842544, + "flow_matching_loss": 0.023458981513977052, + "grad_norm": 0.1727350801229477, + "learning_rate": 2.607608678704859e-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": 0.010793544258922339, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005353818205185235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02694090660661459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023458981513977052, + "weighted_rkd_loss": 0.013470453303307294 + }, + { + "epoch": 2.3115577889447234, + "flow_matching_loss": 0.02446927558630705, + "grad_norm": 0.20639468729496002, + "learning_rate": 2.600894332051139e-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": 0.01072792410850525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047226678114384415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026178515702486038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02446927558630705, + "weighted_rkd_loss": 0.013089257851243019 + }, + { + "epoch": 2.3283082077051924, + "flow_matching_loss": 0.02812897562980652, + "grad_norm": 0.23591165244579315, + "learning_rate": 2.594131828712932e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010303571028634906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005087628937326371, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02569477092474699, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02812897562980652, + "weighted_rkd_loss": 0.012847385462373495 + }, + { + "epoch": 2.3450586264656614, + "flow_matching_loss": 0.023055185005068778, + "grad_norm": 0.2851240336894989, + "learning_rate": 2.5873214645040414e-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": 0.011493522580713034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053280909080058335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028315136022865773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023055185005068778, + "weighted_rkd_loss": 0.014157568011432886 + }, + { + "epoch": 2.3618090452261304, + "flow_matching_loss": 0.023752907849848272, + "grad_norm": 0.2843516170978546, + "learning_rate": 2.5804635373318604e-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": 0.012701206561177968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00591087422799319, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03131328746676445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023752907849848272, + "weighted_rkd_loss": 0.015656643733382224 + }, + { + "epoch": 2.3785594639865995, + "flow_matching_loss": 0.024771238677203655, + "grad_norm": 0.23561623692512512, + "learning_rate": 2.57355834718434e-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": 0.008858569525182247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004208746063522994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02192588523030281, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024771238677203655, + "weighted_rkd_loss": 0.010962942615151405 + }, + { + "epoch": 2.3953098827470685, + "flow_matching_loss": 0.024008012004196645, + "grad_norm": 0.20277637243270874, + "learning_rate": 2.566606196116867e-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": 0.01078841257840395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050114454235881565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0265882708132267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024008012004196645, + "weighted_rkd_loss": 0.01329413540661335 + }, + { + "epoch": 2.4120603015075375, + "flow_matching_loss": 0.02423899285495281, + "grad_norm": 0.21418488025665283, + "learning_rate": 2.559607388239051e-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": 0.010564777301624417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005194381531327963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02632393632084131, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02423899285495281, + "weighted_rkd_loss": 0.013161968160420656 + }, + { + "epoch": 2.4288107202680065, + "flow_matching_loss": 0.021590460650622846, + "grad_norm": 0.23149818181991577, + "learning_rate": 2.552562229701422e-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": 0.010688819037750364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005056107765994966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026433745771646498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021590460650622846, + "weighted_rkd_loss": 0.013216872885823249 + }, + { + "epoch": 2.4455611390284755, + "flow_matching_loss": 0.0240889685228467, + "grad_norm": 0.1726461946964264, + "learning_rate": 2.545471028682036e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009562594909220934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043014321010559796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023426621593534945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0240889685228467, + "weighted_rkd_loss": 0.011713310796767473 + }, + { + "epoch": 2.4623115577889445, + "flow_matching_loss": 0.025627955421805383, + "grad_norm": 0.2549595236778259, + "learning_rate": 2.538334095373e-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": 0.010514754336327314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004727694508619607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025757203251123427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025627955421805383, + "weighted_rkd_loss": 0.012878601625561713 + }, + { + "epoch": 2.4790619765494135, + "flow_matching_loss": 0.024999511800706385, + "grad_norm": 0.20995697379112244, + "learning_rate": 2.531151741966896e-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": 0.00883528389967978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004498904151841998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02216947190463543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024999511800706385, + "weighted_rkd_loss": 0.011084735952317714 + }, + { + "epoch": 2.4958123953098825, + "flow_matching_loss": 0.024381516501307487, + "grad_norm": 0.22541256248950958, + "learning_rate": 2.5239242826431302e-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": 0.009589857840910553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004377200687304139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023556916601955892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024381516501307487, + "weighted_rkd_loss": 0.011778458300977946 + }, + { + "epoch": 2.5125628140703515, + "flow_matching_loss": 0.02520757894963026, + "grad_norm": 0.2073884904384613, + "learning_rate": 2.5166520335541865e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010272404458373786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051223641028627755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02566717304289341, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02520757894963026, + "weighted_rkd_loss": 0.012833586521446705 + }, + { + "epoch": 2.5293132328308205, + "flow_matching_loss": 0.02480332348495722, + "grad_norm": 0.22228868305683136, + "learning_rate": 2.5093353128117996e-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": 0.008547855308279395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004206642787903547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021302353404462338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02480332348495722, + "weighted_rkd_loss": 0.010651176702231169 + }, + { + "epoch": 2.5460636515912896, + "flow_matching_loss": 0.023200121708214283, + "grad_norm": 0.1978033035993576, + "learning_rate": 2.501974440473037e-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": 0.010471200989559292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004926082771271467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025868484750390053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023200121708214283, + "weighted_rkd_loss": 0.012934242375195026 + }, + { + "epoch": 2.5628140703517586, + "flow_matching_loss": 0.02555184094235301, + "grad_norm": 0.22602304816246033, + "learning_rate": 2.4945697385263015e-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": 0.008761562407016754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039977532578632236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021520878095179795, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02555184094235301, + "weighted_rkd_loss": 0.010760439047589897 + }, + { + "epoch": 2.5795644891122276, + "flow_matching_loss": 0.021686122007668018, + "grad_norm": 0.1781231164932251, + "learning_rate": 2.487121530877243e-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": 0.010080217849463224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00466340493876487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024823840521275998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021686122007668018, + "weighted_rkd_loss": 0.012411920260637999 + }, + { + "epoch": 2.5963149078726966, + "flow_matching_loss": 0.022586009837687014, + "grad_norm": 0.21403099596500397, + "learning_rate": 2.4796301433345943e-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": 0.010508734360337257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005213955650106072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02623142395168543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022586009837687014, + "weighted_rkd_loss": 0.013115711975842714 + }, + { + "epoch": 2.6130653266331656, + "flow_matching_loss": 0.023120622523128985, + "grad_norm": 0.18488554656505585, + "learning_rate": 2.4720959035959148e-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": 0.008027114858850836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003637227532453835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019691457133740188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023120622523128985, + "weighted_rkd_loss": 0.009845728566870094 + }, + { + "epoch": 2.6298157453936346, + "flow_matching_loss": 0.02213309984654188, + "grad_norm": 0.2590029537677765, + "learning_rate": 2.464519141233257e-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": 0.008783003455027938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004119690437801182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021685697697103025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02213309984654188, + "weighted_rkd_loss": 0.010842848848551512 + }, + { + "epoch": 2.6465661641541036, + "flow_matching_loss": 0.023288248293101787, + "grad_norm": 0.20496879518032074, + "learning_rate": 2.456900187678754e-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": 0.009698561951518058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004631320876069367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024028444662690163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023288248293101787, + "weighted_rkd_loss": 0.012014222331345081 + }, + { + "epoch": 2.6633165829145726, + "flow_matching_loss": 0.02162441601976752, + "grad_norm": 0.1816551685333252, + "learning_rate": 2.4492393762101147e-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": 0.009497049171477557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004680510167963803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023674608673900366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02162441601976752, + "weighted_rkd_loss": 0.011837304336950183 + }, + { + "epoch": 2.6800670016750416, + "flow_matching_loss": 0.02216140553355217, + "grad_norm": 0.1991337537765503, + "learning_rate": 2.4415370419360508e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009966886276379228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004221933707594871, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024155706353485584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02216140553355217, + "weighted_rkd_loss": 0.012077853176742792 + }, + { + "epoch": 2.6968174204355106, + "flow_matching_loss": 0.02192650279030204, + "grad_norm": 0.2767055034637451, + "learning_rate": 2.4337935217816157e-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": 0.010798801574856043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047045488143339755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02630215175449848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02192650279030204, + "weighted_rkd_loss": 0.01315107587724924 + }, + { + "epoch": 2.7135678391959797, + "flow_matching_loss": 0.02603024896234274, + "grad_norm": 0.20454247295856476, + "learning_rate": 2.4260091544734665e-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": 0.010186881572008134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004692598734982312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025066361762583256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02603024896234274, + "weighted_rkd_loss": 0.012533180881291628 + }, + { + "epoch": 2.7303182579564487, + "flow_matching_loss": 0.025671250931918622, + "grad_norm": 0.20730584859848022, + "learning_rate": 2.4181842805250466e-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": 0.008627374796196819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042751003755256535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02152985017746687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025671250931918622, + "weighted_rkd_loss": 0.010764925088733435 + }, + { + "epoch": 2.7470686767169177, + "flow_matching_loss": 0.02359693106263876, + "grad_norm": 0.188655287027359, + "learning_rate": 2.4103192422216923e-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": 0.010495147341862321, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005308826128020883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02629912067204714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02359693106263876, + "weighted_rkd_loss": 0.01314956033602357 + }, + { + "epoch": 2.7638190954773867, + "flow_matching_loss": 0.023337428458034992, + "grad_norm": 0.23929449915885925, + "learning_rate": 2.402414383605658e-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": 0.010052195470780134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004961214587092399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025065605714917182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023337428458034992, + "weighted_rkd_loss": 0.012532802857458591 + }, + { + "epoch": 2.7805695142378557, + "flow_matching_loss": 0.02219377104192972, + "grad_norm": 0.2548065185546875, + "learning_rate": 2.3944700504610674e-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": 0.010901725804433226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045725111151114104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026375962886959314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02219377104192972, + "weighted_rkd_loss": 0.013187981443479657 + }, + { + "epoch": 2.7973199329983247, + "flow_matching_loss": 0.021554847713559866, + "grad_norm": 0.19051788747310638, + "learning_rate": 2.3864865902987883e-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": 0.009427066240459681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004298139782622457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023152272403240203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021554847713559866, + "weighted_rkd_loss": 0.011576136201620102 + }, + { + "epoch": 2.8140703517587937, + "flow_matching_loss": 0.025536979362368584, + "grad_norm": 0.21837979555130005, + "learning_rate": 2.378464352341231e-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": 0.010489001916721464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051647582557052376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026142761670053007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025536979362368584, + "weighted_rkd_loss": 0.013071380835026503 + }, + { + "epoch": 2.8308207705192627, + "flow_matching_loss": 0.020257471688091756, + "grad_norm": 0.21359702944755554, + "learning_rate": 2.3704036875070728e-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": 0.008906964818015695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004341627052053809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022155556827783585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020257471688091756, + "weighted_rkd_loss": 0.011077778413891792 + }, + { + "epoch": 2.8475711892797317, + "flow_matching_loss": 0.022737276274710894, + "grad_norm": 0.18399080634117126, + "learning_rate": 2.362304948395906e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009278100775554776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00438445636536926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022940658032894135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022737276274710894, + "weighted_rkd_loss": 0.011470329016447067 + }, + { + "epoch": 2.8643216080402008, + "flow_matching_loss": 0.02296969071030617, + "grad_norm": 0.2073369175195694, + "learning_rate": 2.3541684892728157e-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": 0.008697510184720158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004495504498481751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02189052477478981, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02296969071030617, + "weighted_rkd_loss": 0.010945262387394905 + }, + { + "epoch": 2.8810720268006698, + "flow_matching_loss": 0.02656156402081251, + "grad_norm": 0.2169615775346756, + "learning_rate": 2.345994666052883e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009356808802112936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004665453173220158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02337907087057829, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02656156402081251, + "weighted_rkd_loss": 0.011689535435289144 + }, + { + "epoch": 2.8978224455611388, + "flow_matching_loss": 0.02025458849966526, + "grad_norm": 0.19369857013225555, + "learning_rate": 2.337783836285614e-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": 0.009551549004390836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004140716185793281, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023243813868612052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02025458849966526, + "weighted_rkd_loss": 0.011621906934306026 + }, + { + "epoch": 2.914572864321608, + "flow_matching_loss": 0.020585576817393304, + "grad_norm": 0.2022498995065689, + "learning_rate": 2.3295363591393026e-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": 0.008104678569361567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003487278800457716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019696636125445366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020585576817393304, + "weighted_rkd_loss": 0.009848318062722683 + }, + { + "epoch": 2.931323283082077, + "flow_matching_loss": 0.02213856503367424, + "grad_norm": 0.22983857989311218, + "learning_rate": 2.3212525953853174e-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": 0.009665360115468502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004478284833021462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023809005320072175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02213856503367424, + "weighted_rkd_loss": 0.011904502660036087 + }, + { + "epoch": 2.948073701842546, + "flow_matching_loss": 0.02260482348501682, + "grad_norm": 0.2563987076282501, + "learning_rate": 2.3129329073823206e-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": 0.008846775069832802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004327321448363364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02202087165787816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02260482348501682, + "weighted_rkd_loss": 0.01101043582893908 + }, + { + "epoch": 2.964824120603015, + "flow_matching_loss": 0.024081751145422457, + "grad_norm": 0.27132245898246765, + "learning_rate": 2.3045776590604175e-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": 0.011177179915830494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004908815934322774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027263175696134567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024081751145422457, + "weighted_rkd_loss": 0.013631587848067284 + }, + { + "epoch": 2.981574539363484, + "flow_matching_loss": 0.02145145069807768, + "grad_norm": 0.23230737447738647, + "learning_rate": 2.296187215905237e-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": 0.00857270359992981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039184731664136056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021063880529254675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02145145069807768, + "weighted_rkd_loss": 0.010531940264627337 + }, + { + "epoch": 2.998324958123953, + "flow_matching_loss": 0.022610493656247854, + "grad_norm": 0.21160763502120972, + "learning_rate": 2.2877619449419437e-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": 0.00877288361079991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040158325573429465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02156159970909357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022610493656247854, + "weighted_rkd_loss": 0.010780799854546785 + }, + { + "epoch": 3.0150753768844223, + "flow_matching_loss": 0.01961312396451831, + "grad_norm": 0.21592052280902863, + "learning_rate": 2.2793022147191843e-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": 0.008552303677424789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004099966096691787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021204573288559913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01961312396451831, + "weighted_rkd_loss": 0.010602286644279956 + }, + { + "epoch": 3.0318257956448913, + "flow_matching_loss": 0.021133813075721264, + "grad_norm": 0.2408725470304489, + "learning_rate": 2.2708083952929634e-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": 0.009924392122775317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004459633771330118, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024308417923748493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021133813075721264, + "weighted_rkd_loss": 0.012154208961874247 + }, + { + "epoch": 3.0485762144053603, + "flow_matching_loss": 0.021858297660946845, + "grad_norm": 0.21889373660087585, + "learning_rate": 2.2622808582104594e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007935315230861306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036201345268636943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019490764942020178, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021858297660946845, + "weighted_rkd_loss": 0.009745382471010089 + }, + { + "epoch": 3.0653266331658293, + "flow_matching_loss": 0.02066167173907161, + "grad_norm": 0.2489049881696701, + "learning_rate": 2.2537199764937703e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008639211347326637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042247757315635685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021503198333084582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02066167173907161, + "weighted_rkd_loss": 0.010751599166542291 + }, + { + "epoch": 3.0820770519262983, + "flow_matching_loss": 0.021619039587676525, + "grad_norm": 0.18713706731796265, + "learning_rate": 2.245126124623595e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00882501546293497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004201194737106562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021851225569844247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021619039587676525, + "weighted_rkd_loss": 0.010925612784922123 + }, + { + "epoch": 3.0988274706867673, + "flow_matching_loss": 0.0222555180080235, + "grad_norm": 0.16398872435092926, + "learning_rate": 2.2364996785228547e-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": 0.008472857438027859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038555797887966036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020801294595003128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0222555180080235, + "weighted_rkd_loss": 0.010400647297501564 + }, + { + "epoch": 3.1155778894472363, + "flow_matching_loss": 0.02233735201880336, + "grad_norm": 0.195217564702034, + "learning_rate": 2.227841015540248e-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": 0.00930324881337583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003956924844533205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022563422471284865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02233735201880336, + "weighted_rkd_loss": 0.011281711235642432 + }, + { + "epoch": 3.1323283082077054, + "flow_matching_loss": 0.021456127893179654, + "grad_norm": 0.24798332154750824, + "learning_rate": 2.2191505144337433e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00950444871559739, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004383202618919313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023392099887132645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021456127893179654, + "weighted_rkd_loss": 0.011696049943566322 + }, + { + "epoch": 3.1490787269681744, + "flow_matching_loss": 0.020449751988053322, + "grad_norm": 0.21128728985786438, + "learning_rate": 2.210428555354013e-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": 0.007659428799524903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003622169536538422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018941027391701937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020449751988053322, + "weighted_rkd_loss": 0.009470513695850969 + }, + { + "epoch": 3.1658291457286434, + "flow_matching_loss": 0.020440852362662555, + "grad_norm": 0.19786065816879272, + "learning_rate": 2.201675519827802e-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": 0.00879075680859387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00397156581748277, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02155307959765196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020440852362662555, + "weighted_rkd_loss": 0.01077653979882598 + }, + { + "epoch": 3.1825795644891124, + "flow_matching_loss": 0.021140540950000286, + "grad_norm": 0.17299114167690277, + "learning_rate": 2.1928917907412403e-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": 0.008886780729517341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004150486085563898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021924047172069548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021140540950000286, + "weighted_rkd_loss": 0.010962023586034774 + }, + { + "epoch": 3.1993299832495814, + "flow_matching_loss": 0.021827292442321778, + "grad_norm": 0.29607075452804565, + "learning_rate": 2.1840777523230937e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008809508057311177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038985042599961163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021517520304769278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021827292442321778, + "weighted_rkd_loss": 0.010758760152384639 + }, + { + "epoch": 3.2160804020100504, + "flow_matching_loss": 0.02189473919570446, + "grad_norm": 0.21935483813285828, + "learning_rate": 2.1752337901279552e-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": 0.008246973948553204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004177061165682972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02067100889980793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02189473919570446, + "weighted_rkd_loss": 0.010335504449903965 + }, + { + "epoch": 3.2328308207705194, + "flow_matching_loss": 0.02115060854703188, + "grad_norm": 0.21308168768882751, + "learning_rate": 2.166360291019383e-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": 0.008105326350778342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034979252610355615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01970857810229063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02115060854703188, + "weighted_rkd_loss": 0.009854289051145316 + }, + { + "epoch": 3.2495812395309884, + "flow_matching_loss": 0.019409695640206337, + "grad_norm": 0.2219625562429428, + "learning_rate": 2.157457643152973e-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": 0.007863681856542825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037062383722513913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01943360213190317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019409695640206337, + "weighted_rkd_loss": 0.009716801065951585 + }, + { + "epoch": 3.2663316582914574, + "flow_matching_loss": 0.02243496719747782, + "grad_norm": 0.17238378524780273, + "learning_rate": 2.148526235959385e-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": 0.008457799488678575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003480206034146249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020395804755389692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02243496719747782, + "weighted_rkd_loss": 0.010197902377694846 + }, + { + "epoch": 3.2830820770519265, + "flow_matching_loss": 0.021297021582722665, + "grad_norm": 0.21285296976566315, + "learning_rate": 2.1395664601273016e-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": 0.008106729062274099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003673051320947707, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019886509422212838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021297021582722665, + "weighted_rkd_loss": 0.009943254711106419 + }, + { + "epoch": 3.2998324958123955, + "flow_matching_loss": 0.02030411781743169, + "grad_norm": 0.22372476756572723, + "learning_rate": 2.130578707586345e-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": 0.008725496334955096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004435207205824554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021886199712753296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02030411781743169, + "weighted_rkd_loss": 0.010943099856376648 + }, + { + "epoch": 3.3165829145728645, + "flow_matching_loss": 0.021738264430314302, + "grad_norm": 0.20458626747131348, + "learning_rate": 2.1215633714899264e-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": 0.008855098765343428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004159924946725369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021870122756808996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021738264430314302, + "weighted_rkd_loss": 0.010935061378404498 + }, + { + "epoch": 3.3333333333333335, + "flow_matching_loss": 0.022326320968568326, + "grad_norm": 0.20455840229988098, + "learning_rate": 2.112520846198053e-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": 0.008037483179941773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037062566028907896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019781222846359015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022326320968568326, + "weighted_rkd_loss": 0.009890611423179508 + }, + { + "epoch": 3.3500837520938025, + "flow_matching_loss": 0.0228325380012393, + "grad_norm": 0.2642732560634613, + "learning_rate": 2.103451527260074e-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": 0.007869466673582792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033093452220782638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019048278499394656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0228325380012393, + "weighted_rkd_loss": 0.009524139249697328 + }, + { + "epoch": 3.3668341708542715, + "flow_matching_loss": 0.019583611376583578, + "grad_norm": 0.2518172860145569, + "learning_rate": 2.094355811397381e-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": 0.008222047938033938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033477101242169738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0197918058373034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019583611376583578, + "weighted_rkd_loss": 0.0098959029186517 + }, + { + "epoch": 3.3835845896147405, + "flow_matching_loss": 0.024029638990759848, + "grad_norm": 0.19392850995063782, + "learning_rate": 2.0852340964860522e-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": 0.008880104077979923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004434320167638362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022194528207182884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024029638990759848, + "weighted_rkd_loss": 0.011097264103591442 + }, + { + "epoch": 3.4003350083752095, + "flow_matching_loss": 0.02049247771501541, + "grad_norm": 0.1794871985912323, + "learning_rate": 2.076086781539447e-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": 0.007119336817413569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028993004932999613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01713797440752387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02049247771501541, + "weighted_rkd_loss": 0.008568987203761935 + }, + { + "epoch": 3.4170854271356785, + "flow_matching_loss": 0.017801507376134396, + "grad_norm": 0.20647704601287842, + "learning_rate": 2.066914266690755e-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": 0.008224464533850551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00377469330560416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02022362258285284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017801507376134396, + "weighted_rkd_loss": 0.01011181129142642 + }, + { + "epoch": 3.4338358458961475, + "flow_matching_loss": 0.019913967046886683, + "grad_norm": 0.1910584270954132, + "learning_rate": 2.0577169531754906e-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": 0.008146553998813033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033962710993364455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019689378887414934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019913967046886683, + "weighted_rkd_loss": 0.009844689443707467 + }, + { + "epoch": 3.4505862646566166, + "flow_matching_loss": 0.022410364355891943, + "grad_norm": 0.15049278736114502, + "learning_rate": 2.048495243313942e-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": 0.007043796870857477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003043292067013681, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017130885925143956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022410364355891943, + "weighted_rkd_loss": 0.008565442962571978 + }, + { + "epoch": 3.4673366834170856, + "flow_matching_loss": 0.019107151962816715, + "grad_norm": 0.1801232546567917, + "learning_rate": 2.039249540493572e-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": 0.006890717707574367, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027747028158046304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016556138079613446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019107151962816715, + "weighted_rkd_loss": 0.008278069039806723 + }, + { + "epoch": 3.4840871021775546, + "flow_matching_loss": 0.01943407179787755, + "grad_norm": 0.21524561941623688, + "learning_rate": 2.0299802491513755e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007901145797222852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037480345461517574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01955032618716359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01943407179787755, + "weighted_rkd_loss": 0.009775163093581795 + }, + { + "epoch": 3.5008375209380236, + "flow_matching_loss": 0.020493100211024285, + "grad_norm": 0.22612208127975464, + "learning_rate": 2.0206877747561828e-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": 0.0077430652920156716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003920659353025257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019406790286302565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020493100211024285, + "weighted_rkd_loss": 0.009703395143151283 + }, + { + "epoch": 3.5175879396984926, + "flow_matching_loss": 0.017947696149349213, + "grad_norm": 0.21178822219371796, + "learning_rate": 2.011372523790927e-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": 0.008696998190134763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004025266319513321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021419262699782848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017947696149349213, + "weighted_rkd_loss": 0.010709631349891424 + }, + { + "epoch": 3.5343383584589616, + "flow_matching_loss": 0.018908549472689627, + "grad_norm": 0.20278148353099823, + "learning_rate": 2.0020349037348643e-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": 0.008784465352073311, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003880465682595968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02144939647987485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018908549472689627, + "weighted_rkd_loss": 0.010724698239937425 + }, + { + "epoch": 3.5510887772194306, + "flow_matching_loss": 0.018494799360632898, + "grad_norm": 0.1947580724954605, + "learning_rate": 1.9926753230457443e-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": 0.009449501568451524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004069603490643203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022968606650829317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018494799360632898, + "weighted_rkd_loss": 0.011484303325414658 + }, + { + "epoch": 3.5678391959798996, + "flow_matching_loss": 0.02321052923798561, + "grad_norm": 0.21995417773723602, + "learning_rate": 1.983294191141948e-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": 0.007246848475188017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003205891884863377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017699589021503925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02321052923798561, + "weighted_rkd_loss": 0.008849794510751963 + }, + { + "epoch": 3.5845896147403686, + "flow_matching_loss": 0.02240783693268895, + "grad_norm": 0.19936057925224304, + "learning_rate": 1.973891918384575e-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": 0.009624690655618906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004620609059929848, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02386999037116766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02240783693268895, + "weighted_rkd_loss": 0.01193499518558383 + }, + { + "epoch": 3.6013400335008376, + "flow_matching_loss": 0.02322760485112667, + "grad_norm": 0.21677754819393158, + "learning_rate": 1.9644689160594945e-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": 0.0078708321088925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003928346477914601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019670010544359684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02322760485112667, + "weighted_rkd_loss": 0.009835005272179842 + }, + { + "epoch": 3.6180904522613067, + "flow_matching_loss": 0.019495286140590908, + "grad_norm": 0.22844098508358002, + "learning_rate": 1.9550255963593553e-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": 0.007603819016367197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035069640493020416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01871460219845176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019495286140590908, + "weighted_rkd_loss": 0.00935730109922588 + }, + { + "epoch": 3.6348408710217757, + "flow_matching_loss": 0.020784722454845906, + "grad_norm": 0.19066329300403595, + "learning_rate": 1.9455623723655524e-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": 0.006900070700794458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033538440940901635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01715398533269763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020784722454845906, + "weighted_rkd_loss": 0.008576992666348815 + }, + { + "epoch": 3.6515912897822447, + "flow_matching_loss": 0.02208636198192835, + "grad_norm": 0.12610961496829987, + "learning_rate": 1.9360796580301606e-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": 0.006914446130394936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028566413442604245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016685533430427314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02208636198192835, + "weighted_rkd_loss": 0.008342766715213657 + }, + { + "epoch": 3.6683417085427137, + "flow_matching_loss": 0.02097361385822296, + "grad_norm": 0.19243593513965607, + "learning_rate": 1.9265778681578245e-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": 0.007681887224316597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037624599412083627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0191262342967093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02097361385822296, + "weighted_rkd_loss": 0.00956311714835465 + }, + { + "epoch": 3.6850921273031827, + "flow_matching_loss": 0.019382779020816088, + "grad_norm": 0.2503208518028259, + "learning_rate": 1.9170574183876143e-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": 0.008374608866870404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038465243531391026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02059574192389846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019382779020816088, + "weighted_rkd_loss": 0.01029787096194923 + }, + { + "epoch": 3.7018425460636517, + "flow_matching_loss": 0.01804492184892297, + "grad_norm": 0.16802993416786194, + "learning_rate": 1.9075187251748452e-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": 0.009310635272413492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043102250900119545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02293149596080184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01804492184892297, + "weighted_rkd_loss": 0.01146574798040092 + }, + { + "epoch": 3.7185929648241207, + "flow_matching_loss": 0.020559789799153804, + "grad_norm": 0.20331931114196777, + "learning_rate": 1.8979622057728596e-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": 0.007381893135607243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030734885251149533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017837275005877017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020559789799153804, + "weighted_rkd_loss": 0.008918637502938509 + }, + { + "epoch": 3.7353433835845897, + "flow_matching_loss": 0.021128038968890907, + "grad_norm": 0.2322036772966385, + "learning_rate": 1.8883882782147748e-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": 0.007587933773174882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003291653376072645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01846752082929015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021128038968890907, + "weighted_rkd_loss": 0.009233760414645076 + }, + { + "epoch": 3.7520938023450587, + "flow_matching_loss": 0.020190212596207856, + "grad_norm": 0.2527245879173279, + "learning_rate": 1.8787973612951967e-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": 0.008083526277914643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038183052791282536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01998535804450512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020190212596207856, + "weighted_rkd_loss": 0.00999267902225256 + }, + { + "epoch": 3.7688442211055277, + "flow_matching_loss": 0.020125599578022957, + "grad_norm": 0.18612495064735413, + "learning_rate": 1.8691898745519028e-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": 0.007300873147323727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002893792022950947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017495538201183082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020125599578022957, + "weighted_rkd_loss": 0.008747769100591541 + }, + { + "epoch": 3.7855946398659968, + "flow_matching_loss": 0.02288515316322446, + "grad_norm": 0.2038274109363556, + "learning_rate": 1.8595662382474863e-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": 0.007445386983454228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003284679213538766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01817545313388109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02288515316322446, + "weighted_rkd_loss": 0.009087726566940546 + }, + { + "epoch": 3.8023450586264658, + "flow_matching_loss": 0.01813176032155752, + "grad_norm": 0.15173381567001343, + "learning_rate": 1.8499268733509766e-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": 0.007210503658279776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003472392586991191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01789340013638139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01813176032155752, + "weighted_rkd_loss": 0.008946700068190695 + }, + { + "epoch": 3.819095477386935, + "flow_matching_loss": 0.021203831862658264, + "grad_norm": 0.21146638691425323, + "learning_rate": 1.84027220151942e-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": 0.0065764828585088255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003243767865933478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016396733839064835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021203831862658264, + "weighted_rkd_loss": 0.008198366919532418 + }, + { + "epoch": 3.835845896147404, + "flow_matching_loss": 0.02068155473098159, + "grad_norm": 0.22891871631145477, + "learning_rate": 1.83060264507944e-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": 0.008209372824057936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003933777997735888, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02035252321511507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02068155473098159, + "weighted_rkd_loss": 0.010176261607557534 + }, + { + "epoch": 3.852596314907873, + "flow_matching_loss": 0.020051562413573264, + "grad_norm": 0.14797094464302063, + "learning_rate": 1.82091862700876e-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": 0.00664952090010047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002952263760380447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01625130567699671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020051562413573264, + "weighted_rkd_loss": 0.008125652838498355 + }, + { + "epoch": 3.869346733668342, + "flow_matching_loss": 0.02093609981238842, + "grad_norm": 0.2215096801519394, + "learning_rate": 1.8112205709177023e-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": 0.00714435912668705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003198638209141791, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017487356252968312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02093609981238842, + "weighted_rkd_loss": 0.008743678126484156 + }, + { + "epoch": 3.886097152428811, + "flow_matching_loss": 0.019973818585276602, + "grad_norm": 0.18120259046554565, + "learning_rate": 1.8015089010306585e-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": 0.0070549966767430305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003187867905944586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01729786125943065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019973818585276602, + "weighted_rkd_loss": 0.008648930629715324 + }, + { + "epoch": 3.90284757118928, + "flow_matching_loss": 0.019720715191215276, + "grad_norm": 0.22681787610054016, + "learning_rate": 1.7917840421675298e-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": 0.007141135260462761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00324872734490782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017530998028814793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019720715191215276, + "weighted_rkd_loss": 0.008765499014407397 + }, + { + "epoch": 3.919597989949749, + "flow_matching_loss": 0.018794780503958462, + "grad_norm": 0.21289195120334625, + "learning_rate": 1.7820464197251474e-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": 0.006023052660748362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029200530145317314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014966158475726844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018794780503958462, + "weighted_rkd_loss": 0.007483079237863422 + }, + { + "epoch": 3.936348408710218, + "flow_matching_loss": 0.02315736413002014, + "grad_norm": 0.22471879422664642, + "learning_rate": 1.7722964596586627e-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": 0.006645664293318987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028370591229759158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01612838786095381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02315736413002014, + "weighted_rkd_loss": 0.008064193930476905 + }, + { + "epoch": 3.953098827470687, + "flow_matching_loss": 0.019443149212747812, + "grad_norm": 0.19242846965789795, + "learning_rate": 1.7625345884629144e-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": 0.007972657866775989, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036050250404514373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019550340808928014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019443149212747812, + "weighted_rkd_loss": 0.009775170404464007 + }, + { + "epoch": 3.969849246231156, + "flow_matching_loss": 0.02088850997388363, + "grad_norm": 0.1732814460992813, + "learning_rate": 1.7527612331537722e-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": 0.006496066693216562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002922359528020024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015914492681622504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02088850997388363, + "weighted_rkd_loss": 0.007957246340811252 + }, + { + "epoch": 3.986599664991625, + "flow_matching_loss": 0.019626106694340707, + "grad_norm": 0.1980288326740265, + "learning_rate": 1.7429768212494593e-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": 0.006757032917812466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003019305528141558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016533371340483428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019626106694340707, + "weighted_rkd_loss": 0.008266685670241714 + }, + { + "epoch": 4.0033500837520934, + "flow_matching_loss": 0.01870811516419053, + "grad_norm": 0.25888529419898987, + "learning_rate": 1.7331817807518497e-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": 0.008479700377210974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036272499710321426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020586650539189576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01870811516419053, + "weighted_rkd_loss": 0.010293325269594788 + }, + { + "epoch": 4.0201005025125625, + "flow_matching_loss": 0.019916127249598504, + "grad_norm": 0.18142595887184143, + "learning_rate": 1.723376540127745e-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": 0.006715080700814724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028278434183448553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016258004866540433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019916127249598504, + "weighted_rkd_loss": 0.008129002433270216 + }, + { + "epoch": 4.0368509212730315, + "flow_matching_loss": 0.01878450671210885, + "grad_norm": 0.23574140667915344, + "learning_rate": 1.7135615282901357e-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": 0.0064266437198966745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003169252898078412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01602254034951329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01878450671210885, + "weighted_rkd_loss": 0.008011270174756646 + }, + { + "epoch": 4.0536013400335005, + "flow_matching_loss": 0.020211434923112392, + "grad_norm": 0.20220205187797546, + "learning_rate": 1.7037371745794358e-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": 0.006330230925232172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002878256363328546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01553871827200055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020211434923112392, + "weighted_rkd_loss": 0.007769359136000275 + }, + { + "epoch": 4.0703517587939695, + "flow_matching_loss": 0.019851094391196967, + "grad_norm": 0.2334161400794983, + "learning_rate": 1.6939039087447026e-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": 0.0066378221847116945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003039022069424391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016314666345715524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019851094391196967, + "weighted_rkd_loss": 0.008157333172857762 + }, + { + "epoch": 4.0871021775544385, + "flow_matching_loss": 0.020226925425231457, + "grad_norm": 0.1534281224012375, + "learning_rate": 1.6840621609248397e-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": 0.007128366222605109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003230903437361121, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017487635742872952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020226925425231457, + "weighted_rkd_loss": 0.008743817871436476 + }, + { + "epoch": 4.1038525963149075, + "flow_matching_loss": 0.019695611950010062, + "grad_norm": 0.2591371536254883, + "learning_rate": 1.6742123616297794e-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": 0.007243454223498702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00313732223585248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017624230589717627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019695611950010062, + "weighted_rkd_loss": 0.008812115294858813 + }, + { + "epoch": 4.1206030150753765, + "flow_matching_loss": 0.018704015389084817, + "grad_norm": 0.13702625036239624, + "learning_rate": 1.664354941721651e-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": 0.007517511304467916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003035618958529085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018070641439408064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018704015389084817, + "weighted_rkd_loss": 0.009035320719704032 + }, + { + "epoch": 4.1373534338358455, + "flow_matching_loss": 0.017411495372653008, + "grad_norm": 0.19919848442077637, + "learning_rate": 1.6544903323959355e-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": 0.007433804357424379, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003234395873732865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01810200437903404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017411495372653008, + "weighted_rkd_loss": 0.00905100218951702 + }, + { + "epoch": 4.1541038525963145, + "flow_matching_loss": 0.02045935718342662, + "grad_norm": 0.21578043699264526, + "learning_rate": 1.6446189651626023e-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": 0.007500055991113186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003137087286449969, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01813719905912876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02045935718342662, + "weighted_rkd_loss": 0.00906859952956438 + }, + { + "epoch": 4.1708542713567835, + "flow_matching_loss": 0.02151382314041257, + "grad_norm": 0.16690842807292938, + "learning_rate": 1.6347412718272316e-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": 0.0064533458091318606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029357226798310877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01584241455420852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02151382314041257, + "weighted_rkd_loss": 0.00792120727710426 + }, + { + "epoch": 4.187604690117253, + "flow_matching_loss": 0.016593755688518286, + "grad_norm": 0.1668848693370819, + "learning_rate": 1.624857684472131e-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": 0.006326651386916638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00287430175812915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015527604427188636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016593755688518286, + "weighted_rkd_loss": 0.007763802213594318 + }, + { + "epoch": 4.204355108877722, + "flow_matching_loss": 0.018163218908011914, + "grad_norm": 0.21210342645645142, + "learning_rate": 1.6149686354374283e-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": 0.006518031936138868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030372726498171685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016073336359113455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018163218908011914, + "weighted_rkd_loss": 0.008036668179556727 + }, + { + "epoch": 4.221105527638191, + "flow_matching_loss": 0.01736935544759035, + "grad_norm": 0.18703977763652802, + "learning_rate": 1.605074557302165e-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": 0.007781515922397375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003655482176691294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01921851420775056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01736935544759035, + "weighted_rkd_loss": 0.00960925710387528 + }, + { + "epoch": 4.23785594639866, + "flow_matching_loss": 0.02085604052990675, + "grad_norm": 0.1353197544813156, + "learning_rate": 1.59517588286537e-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": 0.007296613347716629, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029753129929304123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0175685397349298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02085604052990675, + "weighted_rkd_loss": 0.0087842698674649 + }, + { + "epoch": 4.254606365159129, + "flow_matching_loss": 0.020376010704785584, + "grad_norm": 0.1928730607032776, + "learning_rate": 1.5852730451271324e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007122312113642693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003080862946808338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017325487174093723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020376010704785584, + "weighted_rkd_loss": 0.008662743587046861 + }, + { + "epoch": 4.271356783919598, + "flow_matching_loss": 0.018556658551096915, + "grad_norm": 0.18790344893932343, + "learning_rate": 1.5753664772696546e-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": 0.006062540132552386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028738549328409134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014998935256153345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018556658551096915, + "weighted_rkd_loss": 0.007499467628076672 + }, + { + "epoch": 4.288107202680067, + "flow_matching_loss": 0.018578741513192654, + "grad_norm": 0.20445463061332703, + "learning_rate": 1.565456612638307e-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": 0.006489685783162713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027945277863182126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015773899294435977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018578741513192654, + "weighted_rkd_loss": 0.007886949647217988 + }, + { + "epoch": 4.304857621440536, + "flow_matching_loss": 0.02046471843495965, + "grad_norm": 0.1576392948627472, + "learning_rate": 1.5555438847226732e-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": 0.006184978596866131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025121036567725243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014882060885429382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02046471843495965, + "weighted_rkd_loss": 0.007441030442714691 + }, + { + "epoch": 4.321608040201005, + "flow_matching_loss": 0.02033666577190161, + "grad_norm": 0.19884538650512695, + "learning_rate": 1.545628727137584e-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": 0.006643129396252334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029675282421521842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016253787092864513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02033666577190161, + "weighted_rkd_loss": 0.008126893546432257 + }, + { + "epoch": 4.338358458961474, + "flow_matching_loss": 0.01932725394144654, + "grad_norm": 0.1970246136188507, + "learning_rate": 1.535711573604153e-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": 0.00804528738372028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036893374170176687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019779912009835244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01932725394144654, + "weighted_rkd_loss": 0.009889956004917622 + }, + { + "epoch": 4.355108877721943, + "flow_matching_loss": 0.01995522724464536, + "grad_norm": 0.1644182652235031, + "learning_rate": 1.5257928579308032e-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": 0.006598327262327075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033264036406762897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016523058246821164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01995522724464536, + "weighted_rkd_loss": 0.008261529123410582 + }, + { + "epoch": 4.371859296482412, + "flow_matching_loss": 0.0177932090125978, + "grad_norm": 0.21884872019290924, + "learning_rate": 1.5158730139942902e-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": 0.008010930195450783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033667325507849453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019388592801988125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0177932090125978, + "weighted_rkd_loss": 0.009694296400994062 + }, + { + "epoch": 4.388609715242881, + "flow_matching_loss": 0.019758843723684548, + "grad_norm": 0.1710134893655777, + "learning_rate": 1.5059524757207239e-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": 0.00648601446300745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029278623638674616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015899891406297682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019758843723684548, + "weighted_rkd_loss": 0.007949945703148841 + }, + { + "epoch": 4.40536013400335, + "flow_matching_loss": 0.017516419384628536, + "grad_norm": 0.15637044608592987, + "learning_rate": 1.4960316770665866e-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": 0.006116408109664917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025824596639722587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014815275836735965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017516419384628536, + "weighted_rkd_loss": 0.007407637918367982 + }, + { + "epoch": 4.422110552763819, + "flow_matching_loss": 0.019074277114123107, + "grad_norm": 0.22107526659965515, + "learning_rate": 1.4861110519997495e-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": 0.006211187061853707, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028650935622863472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015287467464804649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019074277114123107, + "weighted_rkd_loss": 0.007643733732402324 + }, + { + "epoch": 4.438860971524288, + "flow_matching_loss": 0.01782265156507492, + "grad_norm": 0.23777632415294647, + "learning_rate": 1.4761910344804913e-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": 0.007094258349388838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028426593402400614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01703117610886693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01782265156507492, + "weighted_rkd_loss": 0.008515588054433466 + }, + { + "epoch": 4.455611390284757, + "flow_matching_loss": 0.020024440810084344, + "grad_norm": 0.1772475689649582, + "learning_rate": 1.466272058442516e-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": 0.006881606904789806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002932384598534554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016695598419755698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020024440810084344, + "weighted_rkd_loss": 0.008347799209877849 + }, + { + "epoch": 4.472361809045226, + "flow_matching_loss": 0.01836460642516613, + "grad_norm": 0.1673913151025772, + "learning_rate": 1.4563545577739676e-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": 0.0067969394847750665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003281451168004423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016875330172479153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01836460642516613, + "weighted_rkd_loss": 0.008437665086239577 + }, + { + "epoch": 4.489112227805695, + "flow_matching_loss": 0.016627813410013913, + "grad_norm": 0.16218630969524384, + "learning_rate": 1.4464389662984543e-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": 0.006768457312136889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003101827809587121, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01663874238729477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016627813410013913, + "weighted_rkd_loss": 0.008319371193647385 + }, + { + "epoch": 4.505862646566165, + "flow_matching_loss": 0.016898567136377097, + "grad_norm": 0.14560790359973907, + "learning_rate": 1.4365257177560684e-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": 0.005781614547595381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002422610018402338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013985839113593102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016898567136377097, + "weighted_rkd_loss": 0.006992919556796551 + }, + { + "epoch": 4.522613065326633, + "flow_matching_loss": 0.01971247997134924, + "grad_norm": 0.14987064898014069, + "learning_rate": 1.426615245784416e-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": 0.007260903157293797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002818218374159187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01734002446755767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01971247997134924, + "weighted_rkd_loss": 0.008670012233778834 + }, + { + "epoch": 4.539363484087103, + "flow_matching_loss": 0.017527775652706624, + "grad_norm": 0.1931697577238083, + "learning_rate": 1.4167079838996468e-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": 0.007004645606502891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003280059481039643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017289350647479294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017527775652706624, + "weighted_rkd_loss": 0.008644675323739647 + }, + { + "epoch": 4.556113902847571, + "flow_matching_loss": 0.018044522684067487, + "grad_norm": 0.18841363489627838, + "learning_rate": 1.4068043654774897e-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": 0.0072848881827667356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003721295588184148, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0182910718023777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018044522684067487, + "weighted_rkd_loss": 0.00914553590118885 + }, + { + "epoch": 4.572864321608041, + "flow_matching_loss": 0.016927159391343594, + "grad_norm": 0.18175353109836578, + "learning_rate": 1.3969048237342992e-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": 0.006235395558178425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026640282245352863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015134819317609072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016927159391343594, + "weighted_rkd_loss": 0.007567409658804536 + }, + { + "epoch": 4.589614740368509, + "flow_matching_loss": 0.020692834444344044, + "grad_norm": 0.17616070806980133, + "learning_rate": 1.3870097917081011e-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": 0.00688162581063807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029462887556292116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01670954031869769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020692834444344044, + "weighted_rkd_loss": 0.008354770159348846 + }, + { + "epoch": 4.606365159128979, + "flow_matching_loss": 0.01656912127509713, + "grad_norm": 0.17389723658561707, + "learning_rate": 1.3771197022396508e-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": 0.0060537876561284065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002805469243321568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014913044683635236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01656912127509713, + "weighted_rkd_loss": 0.007456522341817618 + }, + { + "epoch": 4.623115577889447, + "flow_matching_loss": 0.018882535118609666, + "grad_norm": 0.1804601401090622, + "learning_rate": 1.3672349879535015e-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": 0.0065174739807844165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026487289811484515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015683677047491074, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018882535118609666, + "weighted_rkd_loss": 0.007841838523745537 + }, + { + "epoch": 4.639865996649917, + "flow_matching_loss": 0.018018264696002005, + "grad_norm": 0.18113377690315247, + "learning_rate": 1.3573560812390781e-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": 0.006505385902710259, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00266862609423697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015679398085922003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018018264696002005, + "weighted_rkd_loss": 0.007839699042961001 + }, + { + "epoch": 4.656616415410385, + "flow_matching_loss": 0.017154605966061352, + "grad_norm": 0.22344084084033966, + "learning_rate": 1.3474834142317635e-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": 0.00781757920049131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003766122250817716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019401280861347913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017154605966061352, + "weighted_rkd_loss": 0.009700640430673957 + }, + { + "epoch": 4.673366834170855, + "flow_matching_loss": 0.020884703472256662, + "grad_norm": 0.16884377598762512, + "learning_rate": 1.3376174187939955e-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": 0.00663802158087492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002596668968908489, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015872712060809136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020884703472256662, + "weighted_rkd_loss": 0.007936356030404568 + }, + { + "epoch": 4.690117252931323, + "flow_matching_loss": 0.01987990001216531, + "grad_norm": 0.17503158748149872, + "learning_rate": 1.3277585264963761e-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": 0.006741104461252689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00272078113630414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01620299005880952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01987990001216531, + "weighted_rkd_loss": 0.00810149502940476 + }, + { + "epoch": 4.706867671691793, + "flow_matching_loss": 0.018524412903934716, + "grad_norm": 0.2198275923728943, + "learning_rate": 1.3179071685987928e-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": 0.006492939265444875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026159399072639643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015601818356662989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018524412903934716, + "weighted_rkd_loss": 0.0078009091783314945 + }, + { + "epoch": 4.723618090452261, + "flow_matching_loss": 0.018392641749233007, + "grad_norm": 0.15910054743289948, + "learning_rate": 1.308063776031554e-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": 0.006039766082540155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022717000334523617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01435123234987259, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018392641749233007, + "weighted_rkd_loss": 0.007175616174936295 + }, + { + "epoch": 4.740368509212731, + "flow_matching_loss": 0.016925379633903503, + "grad_norm": 0.1666567623615265, + "learning_rate": 1.2982287793765389e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005827722325921059, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002627475606277585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014282920211553574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016925379633903503, + "weighted_rkd_loss": 0.007141460105776787 + }, + { + "epoch": 4.757118927973199, + "flow_matching_loss": 0.016931863874197005, + "grad_norm": 0.16195693612098694, + "learning_rate": 1.2884026088483616e-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": 0.006095975963398814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00292929713614285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015121249109506607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016931863874197005, + "weighted_rkd_loss": 0.007560624554753303 + }, + { + "epoch": 4.773869346733669, + "flow_matching_loss": 0.02151737492531538, + "grad_norm": 0.16743698716163635, + "learning_rate": 1.2785856942755543e-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": 0.006752799730747938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033896766020916403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016895275842398406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02151737492531538, + "weighted_rkd_loss": 0.008447637921199203 + }, + { + "epoch": 4.790619765494137, + "flow_matching_loss": 0.018972883094102146, + "grad_norm": 0.21053996682167053, + "learning_rate": 1.2687784650817616e-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": 0.006391771929338574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025758426752872767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015359386708587408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018972883094102146, + "weighted_rkd_loss": 0.007679693354293704 + }, + { + "epoch": 4.807370184254607, + "flow_matching_loss": 0.020155242085456847, + "grad_norm": 0.15815846621990204, + "learning_rate": 1.2589813502669603e-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": 0.0053243459668010475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020828464068472384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012731538340449332, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020155242085456847, + "weighted_rkd_loss": 0.006365769170224666 + }, + { + "epoch": 4.824120603015075, + "flow_matching_loss": 0.019101917929947376, + "grad_norm": 0.14777660369873047, + "learning_rate": 1.2491947783886907e-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": 0.006484691519290209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031254583853296937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016094841342419387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019101917929947376, + "weighted_rkd_loss": 0.008047420671209693 + }, + { + "epoch": 4.840871021775545, + "flow_matching_loss": 0.01791600426658988, + "grad_norm": 0.12246904522180557, + "learning_rate": 1.2394191775433106e-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": 0.005786749953404069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028565174899995325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014430017489939929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01791600426658988, + "weighted_rkd_loss": 0.007215008744969964 + }, + { + "epoch": 4.857621440536013, + "flow_matching_loss": 0.016301574185490608, + "grad_norm": 0.1843765676021576, + "learning_rate": 1.2296549753472696e-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": 0.005566553748212755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026679943315684796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013801102060824633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016301574185490608, + "weighted_rkd_loss": 0.006900551030412316 + }, + { + "epoch": 4.874371859296483, + "flow_matching_loss": 0.01598440548405051, + "grad_norm": 0.17817851901054382, + "learning_rate": 1.2199025989184032e-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": 0.00497903733048588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021776941372081637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012135768868029117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01598440548405051, + "weighted_rkd_loss": 0.006067884434014559 + }, + { + "epoch": 4.891122278056951, + "flow_matching_loss": 0.018774052895605565, + "grad_norm": 0.13188794255256653, + "learning_rate": 1.21016247485725e-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": 0.005346168461255729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025814786669798194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013273815717548132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018774052895605565, + "weighted_rkd_loss": 0.006636907858774066 + }, + { + "epoch": 4.907872696817421, + "flow_matching_loss": 0.018068937305361032, + "grad_norm": 0.1632096916437149, + "learning_rate": 1.2004350292283897e-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": 0.005146085587330163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023135139606893064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012605685275048018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018068937305361032, + "weighted_rkd_loss": 0.006302842637524009 + }, + { + "epoch": 4.924623115577889, + "flow_matching_loss": 0.017883090209215878, + "grad_norm": 0.15789315104484558, + "learning_rate": 1.1907206875418068e-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": 0.006365909287706017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027634695288725197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01549528781324625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017883090209215878, + "weighted_rkd_loss": 0.007747643906623125 + }, + { + "epoch": 4.941373534338359, + "flow_matching_loss": 0.018130379170179366, + "grad_norm": 0.18408702313899994, + "learning_rate": 1.1810198747342769e-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": 0.0056016219314187765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025636515463702382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01376689551398158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018130379170179366, + "weighted_rkd_loss": 0.00688344775699079 + }, + { + "epoch": 4.958123953098827, + "flow_matching_loss": 0.015793959517031907, + "grad_norm": 0.1873013824224472, + "learning_rate": 1.1713330151507786e-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": 0.006022406509146094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025855567771941425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014630369748920203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015793959517031907, + "weighted_rkd_loss": 0.007315184874460101 + }, + { + "epoch": 4.974874371859297, + "flow_matching_loss": 0.01843378497287631, + "grad_norm": 0.1450514942407608, + "learning_rate": 1.161660532525931e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063908693846315145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002764756348915398, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015546495281159877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01843378497287631, + "weighted_rkd_loss": 0.0077732476405799385 + }, + { + "epoch": 4.991624790619765, + "flow_matching_loss": 0.015466341562569142, + "grad_norm": 0.17179778218269348, + "learning_rate": 1.1520028499654584e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060694361571222545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024145430303178727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014553415589034557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015466341562569142, + "weighted_rkd_loss": 0.007276707794517279 + }, + { + "epoch": 5.008375209380235, + "flow_matching_loss": 0.017944001685827972, + "grad_norm": 0.21506550908088684, + "learning_rate": 1.1423603899276833e-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": 0.006832077959552407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002628091024234891, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016292246989905834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017944001685827972, + "weighted_rkd_loss": 0.008146123494952917 + }, + { + "epoch": 5.025125628140704, + "flow_matching_loss": 0.017549295723438264, + "grad_norm": 0.26095050573349, + "learning_rate": 1.1327335742050442e-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": 0.006050047976896167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002698522701393813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01479861866682768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017549295723438264, + "weighted_rkd_loss": 0.00739930933341384 + }, + { + "epoch": 5.041876046901173, + "flow_matching_loss": 0.016273344960063697, + "grad_norm": 0.1535094678401947, + "learning_rate": 1.1231228239056467e-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": 0.005975645314902067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023796642548404635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014330955035984517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016273344960063697, + "weighted_rkd_loss": 0.007165477517992258 + }, + { + "epoch": 5.058626465661642, + "flow_matching_loss": 0.018688787426799535, + "grad_norm": 0.1590380221605301, + "learning_rate": 1.113528559434845e-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": 0.005144167877733708, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002109654329251498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012397990189492702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018688787426799535, + "weighted_rkd_loss": 0.006198995094746351 + }, + { + "epoch": 5.075376884422111, + "flow_matching_loss": 0.020107445027679204, + "grad_norm": 0.2060288041830063, + "learning_rate": 1.1039512004768476e-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": 0.006088935793377459, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002645562624093145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014823434315621853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020107445027679204, + "weighted_rkd_loss": 0.007411717157810927 + }, + { + "epoch": 5.09212730318258, + "flow_matching_loss": 0.017590080946683885, + "grad_norm": 0.1265738606452942, + "learning_rate": 1.0943911659763612e-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": 0.006028936221264302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025103176943957807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014568190462887287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017590080946683885, + "weighted_rkd_loss": 0.007284095231443644 + }, + { + "epoch": 5.108877721943049, + "flow_matching_loss": 0.017241567280143498, + "grad_norm": 0.15954196453094482, + "learning_rate": 1.084848874120266e-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": 0.006110600777901709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026879373588599266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01490913899615407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017241567280143498, + "weighted_rkd_loss": 0.007454569498077035 + }, + { + "epoch": 5.125628140703517, + "flow_matching_loss": 0.018878405541181566, + "grad_norm": 0.17765270173549652, + "learning_rate": 1.0753247423193196e-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": 0.00597507394850254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027689284761436284, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014719076454639435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018878405541181566, + "weighted_rkd_loss": 0.007359538227319717 + }, + { + "epoch": 5.142378559463987, + "flow_matching_loss": 0.017817400768399237, + "grad_norm": 0.17209725081920624, + "learning_rate": 1.0658191871899012e-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": 0.005254556564614177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024205024936236443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012929615657776594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017817400768399237, + "weighted_rkd_loss": 0.006464807828888297 + }, + { + "epoch": 5.159128978224456, + "flow_matching_loss": 0.01816714182496071, + "grad_norm": 0.14104831218719482, + "learning_rate": 1.0563326245357852e-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": 0.0050902100512757896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002153707609977573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012334127770736813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01816714182496071, + "weighted_rkd_loss": 0.006167063885368407 + }, + { + "epoch": 5.175879396984925, + "flow_matching_loss": 0.01670421687886119, + "grad_norm": 0.1472315937280655, + "learning_rate": 1.0468654693299557e-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": 0.0054107870440930125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027880519977770747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013609626144170762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01670421687886119, + "weighted_rkd_loss": 0.006804813072085381 + }, + { + "epoch": 5.192629815745394, + "flow_matching_loss": 0.015585028752684593, + "grad_norm": 0.24463646113872528, + "learning_rate": 1.0374181356964501e-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": 0.006263877777382731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026189771597273648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015146732609719039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015585028752684593, + "weighted_rkd_loss": 0.007573366304859519 + }, + { + "epoch": 5.209380234505863, + "flow_matching_loss": 0.016866875905543567, + "grad_norm": 0.20484580099582672, + "learning_rate": 1.0279910368922465e-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": 0.005633864691480995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025424073450267316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013810136541724204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016866875905543567, + "weighted_rkd_loss": 0.006905068270862102 + }, + { + "epoch": 5.226130653266332, + "flow_matching_loss": 0.015790032222867012, + "grad_norm": 0.1716005951166153, + "learning_rate": 1.018584585289186e-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": 0.005199952656403184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00229923086008057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012699136231094598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015790032222867012, + "weighted_rkd_loss": 0.006349568115547299 + }, + { + "epoch": 5.242881072026801, + "flow_matching_loss": 0.017556129675358533, + "grad_norm": 0.22674204409122467, + "learning_rate": 1.0091991923559348e-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": 0.005729359481483698, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024778502178378404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013936569122597576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017556129675358533, + "weighted_rkd_loss": 0.006968284561298788 + }, + { + "epoch": 5.259631490787269, + "flow_matching_loss": 0.017610132601112126, + "grad_norm": 0.14236882328987122, + "learning_rate": 9.99835268639984e-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": 0.00562389271799475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023109326721169055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01355871818959713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017610132601112126, + "weighted_rkd_loss": 0.006779359094798565 + }, + { + "epoch": 5.276381909547739, + "flow_matching_loss": 0.015706593077629803, + "grad_norm": 0.1687908172607422, + "learning_rate": 9.904932237496913e-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": 0.00528958635404706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022576195537112654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012836792226880788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015706593077629803, + "weighted_rkd_loss": 0.006418396113440394 + }, + { + "epoch": 5.293132328308207, + "flow_matching_loss": 0.014574210066348315, + "grad_norm": 0.16634498536586761, + "learning_rate": 9.811734663363659e-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": 0.005389407277107239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002465158875565976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013243973534554244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014574210066348315, + "weighted_rkd_loss": 0.006621986767277122 + }, + { + "epoch": 5.309882747068677, + "flow_matching_loss": 0.01800208482891321, + "grad_norm": 0.197093203663826, + "learning_rate": 9.718764040763875e-06, + "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": 0.006899569206871093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002856101421639323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01665523974224925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01800208482891321, + "weighted_rkd_loss": 0.008327619871124625 + }, + { + "epoch": 5.326633165829146, + "flow_matching_loss": 0.015227837394922972, + "grad_norm": 0.18592070043087006, + "learning_rate": 9.626024436533777e-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": 0.006064104102551937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002545801061205566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014674009475857019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015227837394922972, + "weighted_rkd_loss": 0.007337004737928509 + }, + { + "epoch": 5.343383584589615, + "flow_matching_loss": 0.018127095606178045, + "grad_norm": 0.17893438041210175, + "learning_rate": 9.533519907404095e-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": 0.005131047265604138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023227728554047642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012584867607802152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018127095606178045, + "weighted_rkd_loss": 0.006292433803901076 + }, + { + "epoch": 5.360134003350084, + "flow_matching_loss": 0.01897403774783015, + "grad_norm": 0.1659892350435257, + "learning_rate": 9.4412544998226e-06, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005119132879190147, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022993266000412406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012537592463195324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01897403774783015, + "weighted_rkd_loss": 0.006268796231597662 + }, + { + "epoch": 5.376884422110553, + "flow_matching_loss": 0.017207826394587754, + "grad_norm": 0.15151554346084595, + "learning_rate": 9.349232249777124e-06, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061891798861324785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025767922634258866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014955152198672295, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017207826394587754, + "weighted_rkd_loss": 0.0074775760993361475 + }, + { + "epoch": 5.393634840871022, + "flow_matching_loss": 0.018857961520552635, + "grad_norm": 0.2096855342388153, + "learning_rate": 9.257457182618985e-06, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005416905065067113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002632979757618159, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01346679013222456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018857961520552635, + "weighted_rkd_loss": 0.00673339506611228 + }, + { + "epoch": 5.410385259631491, + "flow_matching_loss": 0.01634107967838645, + "grad_norm": 0.20736907422542572, + "learning_rate": 9.165933312886944e-06, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065269402228295805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027695797849446534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015823460090905427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01634107967838645, + "weighted_rkd_loss": 0.007911730045452713 + }, + { + "epoch": 5.42713567839196, + "flow_matching_loss": 0.016097908932715656, + "grad_norm": 0.15641118586063385, + "learning_rate": 9.07466464413155e-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": 0.0045119600370526316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002176347211934626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011200267216190695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016097908932715656, + "weighted_rkd_loss": 0.005600133608095348 + }, + { + "epoch": 5.443886097152429, + "flow_matching_loss": 0.018399234488606454, + "grad_norm": 0.20723454654216766, + "learning_rate": 8.983655168740048e-06, + "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": 0.005364138376899063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022491870215162633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01297746361233294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018399234488606454, + "weighted_rkd_loss": 0.00648873180616647 + }, + { + "epoch": 5.460636515912898, + "flow_matching_loss": 0.016172559838742016, + "grad_norm": 0.17914651334285736, + "learning_rate": 8.892908867761718e-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": 0.004746598890051246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020948843797668816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01158808209002018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016172559838742016, + "weighted_rkd_loss": 0.00579404104501009 + }, + { + "epoch": 5.477386934673367, + "flow_matching_loss": 0.01892400048673153, + "grad_norm": 0.1844613254070282, + "learning_rate": 8.802429710733745e-06, + "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": 0.004971841280348599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020754300872795283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012019112706184387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01892400048673153, + "weighted_rkd_loss": 0.006009556353092194 + }, + { + "epoch": 5.494137353433836, + "flow_matching_loss": 0.017203427013009788, + "grad_norm": 0.14044742286205292, + "learning_rate": 8.71222165550757e-06, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005311322677880525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002102222468238324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012724867742508649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017203427013009788, + "weighted_rkd_loss": 0.0063624338712543246 + }, + { + "epoch": 5.510887772194305, + "flow_matching_loss": 0.017448701057583095, + "grad_norm": 0.13229121267795563, + "learning_rate": 8.622288648075748e-06, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004944969573989511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022162436041980984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01210618270561099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017448701057583095, + "weighted_rkd_loss": 0.006053091352805495 + }, + { + "epoch": 5.527638190954773, + "flow_matching_loss": 0.018544258456677197, + "grad_norm": 0.18051110208034515, + "learning_rate": 8.532634622399391e-06, + "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": 0.005711289774626494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021691814879886808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013591760955750942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018544258456677197, + "weighted_rkd_loss": 0.006795880477875471 + }, + { + "epoch": 5.544388609715243, + "flow_matching_loss": 0.01410124646499753, + "grad_norm": 0.11878155916929245, + "learning_rate": 8.44326350023601e-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": 0.004789017117582261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019207892357371748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011498823389410972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01410124646499753, + "weighted_rkd_loss": 0.005749411694705486 + }, + { + "epoch": 5.561139028475711, + "flow_matching_loss": 0.016507032420486213, + "grad_norm": 0.1763717085123062, + "learning_rate": 8.354179190968002e-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": 0.005711464490741491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002308214281219989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01373114325106144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016507032420486213, + "weighted_rkd_loss": 0.00686557162553072 + }, + { + "epoch": 5.577889447236181, + "flow_matching_loss": 0.020039274729788302, + "grad_norm": 0.13657470047473907, + "learning_rate": 8.265385591431661e-06, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005672144400887191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025627062539570035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013906995020806789, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020039274729788302, + "weighted_rkd_loss": 0.006953497510403394 + }, + { + "epoch": 5.594639865996649, + "flow_matching_loss": 0.015484216436743737, + "grad_norm": 0.1777724176645279, + "learning_rate": 8.176886585746671e-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": 0.006078959535807371, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025322111207060514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014690130297094584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015484216436743737, + "weighted_rkd_loss": 0.007345065148547292 + }, + { + "epoch": 5.611390284757119, + "flow_matching_loss": 0.01517918286845088, + "grad_norm": 0.16232475638389587, + "learning_rate": 8.088686045146247e-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": 0.004736245633102954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024781554588116706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011950646806508303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01517918286845088, + "weighted_rkd_loss": 0.005975323403254152 + }, + { + "epoch": 5.628140703517588, + "flow_matching_loss": 0.015617287997156382, + "grad_norm": 0.160396009683609, + "learning_rate": 8.00078782780775e-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": 0.004729195917025209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002444211218971759, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011902603134512902, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015617287997156382, + "weighted_rkd_loss": 0.005951301567256451 + }, + { + "epoch": 5.644891122278057, + "flow_matching_loss": 0.019811277277767658, + "grad_norm": 0.1667182594537735, + "learning_rate": 7.913195778683979e-06, + "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": 0.004932631156407297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002067512331996113, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011932774633169174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019811277277767658, + "weighted_rkd_loss": 0.005966387316584587 + }, + { + "epoch": 5.661641541038526, + "flow_matching_loss": 0.016531050205230713, + "grad_norm": 0.15772488713264465, + "learning_rate": 7.825913729334908e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00438170931302011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018925598124042153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0106559784617275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016531050205230713, + "weighted_rkd_loss": 0.00532798923086375 + }, + { + "epoch": 5.678391959798995, + "flow_matching_loss": 0.018603511899709702, + "grad_norm": 0.1836947351694107, + "learning_rate": 7.73894549776013e-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": 0.004417502367869019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019307143054902553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01076571922749281, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018603511899709702, + "weighted_rkd_loss": 0.005382859613746405 + }, + { + "epoch": 5.695142378559464, + "flow_matching_loss": 0.0157127789221704, + "grad_norm": 0.15395602583885193, + "learning_rate": 7.65229488823184e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005178571795113385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020820079836994408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012439151760190725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0157127789221704, + "weighted_rkd_loss": 0.006219575880095363 + }, + { + "epoch": 5.711892797319933, + "flow_matching_loss": 0.01738949567079544, + "grad_norm": 0.22008782625198364, + "learning_rate": 7.565965691128396e-06, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005835405667312443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022089136065915225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01387972510419786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01738949567079544, + "weighted_rkd_loss": 0.00693986255209893 + }, + { + "epoch": 5.728643216080402, + "flow_matching_loss": 0.015781704988330602, + "grad_norm": 0.17194882035255432, + "learning_rate": 7.47996168276856e-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": 0.004093670891597867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015222638729028403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009709605574607849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015781704988330602, + "weighted_rkd_loss": 0.004854802787303924 + }, + { + "epoch": 5.745393634840871, + "flow_matching_loss": 0.01642342172563076, + "grad_norm": 0.14569294452667236, + "learning_rate": 7.394286625246258e-06, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00585379155818373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002785720257088542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014493303652852774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01642342172563076, + "weighted_rkd_loss": 0.007246651826426387 + }, + { + "epoch": 5.76214405360134, + "flow_matching_loss": 0.016525352746248244, + "grad_norm": 0.1307617723941803, + "learning_rate": 7.308944266266063e-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": 0.005043041077442467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002132825099397451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012218907242640853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016525352746248244, + "weighted_rkd_loss": 0.0061094536213204265 + }, + { + "epoch": 5.778894472361809, + "flow_matching_loss": 0.016573724709451197, + "grad_norm": 0.13970951735973358, + "learning_rate": 7.223938338979233e-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": 0.0046006364049389955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001974192395573482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011175465257838368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016573724709451197, + "weighted_rkd_loss": 0.005587732628919184 + }, + { + "epoch": 5.795644891122278, + "flow_matching_loss": 0.018552876077592373, + "grad_norm": 0.16639065742492676, + "learning_rate": 7.139272561820397e-06, + "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": 0.005241074599325657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020407545613124966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012522903643548489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018552876077592373, + "weighted_rkd_loss": 0.0062614518217742445 + }, + { + "epoch": 5.812395309882747, + "flow_matching_loss": 0.014944805391132831, + "grad_norm": 0.18595440685749054, + "learning_rate": 7.054950638344936e-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": 0.004425915516912937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00172965683741495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010581487929448486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014944805391132831, + "weighted_rkd_loss": 0.005290743964724243 + }, + { + "epoch": 5.8291457286432165, + "flow_matching_loss": 0.016345801949501037, + "grad_norm": 0.1472901552915573, + "learning_rate": 6.970976257066941e-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": 0.0051440172363072635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020479016355238857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012335935980081559, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016345801949501037, + "weighted_rkd_loss": 0.006167967990040779 + }, + { + "epoch": 5.8458961474036855, + "flow_matching_loss": 0.01786613268777728, + "grad_norm": 0.1330331712961197, + "learning_rate": 6.887353091297896e-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": 0.00458008588757366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001994237513281405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011154409311711789, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01786613268777728, + "weighted_rkd_loss": 0.0055772046558558944 + }, + { + "epoch": 5.8626465661641545, + "flow_matching_loss": 0.01581496661528945, + "grad_norm": 0.1423383206129074, + "learning_rate": 6.804084798985965e-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": 0.004506601137109101, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019407646264880895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010953966854140162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01581496661528945, + "weighted_rkd_loss": 0.005476983427070081 + }, + { + "epoch": 5.8793969849246235, + "flow_matching_loss": 0.017252940963953734, + "grad_norm": 0.20213234424591064, + "learning_rate": 6.721175022556009e-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": 0.004888100852258504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019750437000766396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01175124542787671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017252940963953734, + "weighted_rkd_loss": 0.005875622713938355 + }, + { + "epoch": 5.8961474036850925, + "flow_matching_loss": 0.016892566438764335, + "grad_norm": 0.17500615119934082, + "learning_rate": 6.638627388750245e-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": 0.005072918278165162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002193704899400473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012339541455730796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016892566438764335, + "weighted_rkd_loss": 0.006169770727865398 + }, + { + "epoch": 5.9128978224455615, + "flow_matching_loss": 0.018169266637414694, + "grad_norm": 0.1558530330657959, + "learning_rate": 6.556445508469583e-06, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004949927981942892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019603127497248352, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01186016877181828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018169266637414694, + "weighted_rkd_loss": 0.00593008438590914 + }, + { + "epoch": 5.9296482412060305, + "flow_matching_loss": 0.01648208061233163, + "grad_norm": 0.13735370337963104, + "learning_rate": 6.474632976615708e-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": 0.004612542688846588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002068738080561161, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01129382341168821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01648208061233163, + "weighted_rkd_loss": 0.005646911705844105 + }, + { + "epoch": 5.9463986599664995, + "flow_matching_loss": 0.016126606985926627, + "grad_norm": 0.15549711883068085, + "learning_rate": 6.393193371933791e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004885917692445218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001710111042484641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011481946427375078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016126606985926627, + "weighted_rkd_loss": 0.005740973213687539 + }, + { + "epoch": 5.9631490787269685, + "flow_matching_loss": 0.015177852846682072, + "grad_norm": 0.1654539257287979, + "learning_rate": 6.312130256855977e-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": 0.004428159282542765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001902568421792239, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010758886951953173, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015177852846682072, + "weighted_rkd_loss": 0.0053794434759765865 + }, + { + "epoch": 5.9798994974874375, + "flow_matching_loss": 0.015710619650781153, + "grad_norm": 0.1425623595714569, + "learning_rate": 6.23144717734552e-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": 0.003932594321668148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016364209004677832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009501609532162548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015710619650781153, + "weighted_rkd_loss": 0.004750804766081274 + }, + { + "epoch": 5.9966499162479066, + "flow_matching_loss": 0.017744600120931864, + "grad_norm": 0.20558246970176697, + "learning_rate": 6.151147662741695e-06, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051352317910641435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002128293039277196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01239875671453774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017744600120931864, + "weighted_rkd_loss": 0.00619937835726887 + }, + { + "epoch": 6.013400335008376, + "flow_matching_loss": 0.016692350897938014, + "grad_norm": 0.15531615912914276, + "learning_rate": 6.071235225605415e-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": 0.0044694194104522465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018079525209032, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010746791213750839, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016692350897938014, + "weighted_rkd_loss": 0.005373395606875419 + }, + { + "epoch": 6.030150753768845, + "flow_matching_loss": 0.015450216457247734, + "grad_norm": 0.17744119465351105, + "learning_rate": 5.991713361565546e-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": 0.004569414188154042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018782433704473077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011017071735113858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015450216457247734, + "weighted_rkd_loss": 0.005508535867556929 + }, + { + "epoch": 6.046901172529314, + "flow_matching_loss": 0.015183007437735796, + "grad_norm": 0.21500062942504883, + "learning_rate": 5.912585549166051e-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": 0.004672595020383597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019326795474626124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011277869530022144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015183007437735796, + "weighted_rkd_loss": 0.005638934765011072 + }, + { + "epoch": 6.063651591289783, + "flow_matching_loss": 0.01673786323517561, + "grad_norm": 0.15154196321964264, + "learning_rate": 5.833855249713777e-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": 0.004518775991164148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00174813725752756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010785689298063516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01673786323517561, + "weighted_rkd_loss": 0.005392844649031758 + }, + { + "epoch": 6.080402010050252, + "flow_matching_loss": 0.016367413848638535, + "grad_norm": 0.15717095136642456, + "learning_rate": 5.7555259071270645e-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": 0.004585496545769274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002205381088424474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01137637412175536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016367413848638535, + "weighted_rkd_loss": 0.00568818706087768 + }, + { + "epoch": 6.097152428810721, + "flow_matching_loss": 0.015005042776465416, + "grad_norm": 0.13135921955108643, + "learning_rate": 5.677600947785117e-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": 0.003971590660512447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001454745524097234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009397926880046726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015005042776465416, + "weighted_rkd_loss": 0.004698963440023363 + }, + { + "epoch": 6.11390284757119, + "flow_matching_loss": 0.01685640923678875, + "grad_norm": 0.18737980723381042, + "learning_rate": 5.600083780378099e-06, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005439688265323639, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002393451309762895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013272827863693238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01685640923678875, + "weighted_rkd_loss": 0.006636413931846619 + }, + { + "epoch": 6.130653266331659, + "flow_matching_loss": 0.016260798461735247, + "grad_norm": 0.13233205676078796, + "learning_rate": 5.522977795758038e-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": 0.005148716387338936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002030079159885645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012327512074261904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016260798461735247, + "weighted_rkd_loss": 0.006163756037130952 + }, + { + "epoch": 6.147403685092128, + "flow_matching_loss": 0.0165544499643147, + "grad_norm": 0.19198189675807953, + "learning_rate": 5.4462863667904805e-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": 0.004077224410139024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020296421251259746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01018409114331007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0165544499643147, + "weighted_rkd_loss": 0.005092045571655035 + }, + { + "epoch": 6.164154103852597, + "flow_matching_loss": 0.017229913640767335, + "grad_norm": 0.1838519275188446, + "learning_rate": 5.370012848206983e-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": 0.004221946839243174, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017063068808056414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010150200640782713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017229913640767335, + "weighted_rkd_loss": 0.005075100320391357 + }, + { + "epoch": 6.180904522613066, + "flow_matching_loss": 0.016627094615250824, + "grad_norm": 0.14189720153808594, + "learning_rate": 5.29416057645834e-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": 0.004839239851571619, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002113574917893857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011792054679244757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016627094615250824, + "weighted_rkd_loss": 0.005896027339622379 + }, + { + "epoch": 6.197654941373535, + "flow_matching_loss": 0.016476182825863363, + "grad_norm": 0.14053557813167572, + "learning_rate": 5.218732869568635e-06, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00519660187419504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023840540554374455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012777257803827524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016476182825863363, + "weighted_rkd_loss": 0.006388628901913762 + }, + { + "epoch": 6.214405360134004, + "flow_matching_loss": 0.015468492917716504, + "grad_norm": 0.1435490995645523, + "learning_rate": 5.14373302699013e-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": 0.004912928817793727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018873865308705717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01171324416063726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015468492917716504, + "weighted_rkd_loss": 0.00585662208031863 + }, + { + "epoch": 6.231155778894473, + "flow_matching_loss": 0.015904262941330672, + "grad_norm": 0.15239299833774567, + "learning_rate": 5.069164329458894e-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": 0.004692935664206743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018197374884039164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011205608956515789, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015904262941330672, + "weighted_rkd_loss": 0.0056028044782578945 + }, + { + "epoch": 6.247906197654942, + "flow_matching_loss": 0.017712344974279405, + "grad_norm": 0.17586107552051544, + "learning_rate": 4.995030038851342e-06, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004220141563564539, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018938871507998556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010334170144051313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017712344974279405, + "weighted_rkd_loss": 0.005167085072025656 + }, + { + "epoch": 6.264656616415411, + "flow_matching_loss": 0.015475019440054893, + "grad_norm": 0.18944132328033447, + "learning_rate": 4.9213333980415e-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": 0.004483212856575847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001734148757532239, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010700574517250061, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015475019440054893, + "weighted_rkd_loss": 0.005350287258625031 + }, + { + "epoch": 6.28140703517588, + "flow_matching_loss": 0.016900824569165706, + "grad_norm": 0.16883336007595062, + "learning_rate": 4.848077630759175e-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": 0.00466771477367729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019422173150815069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011277646850794553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016900824569165706, + "weighted_rkd_loss": 0.005638823425397277 + }, + { + "epoch": 6.298157453936349, + "flow_matching_loss": 0.015882111340761184, + "grad_norm": 0.1701260656118393, + "learning_rate": 4.775265941448951e-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": 0.004563337680883706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017544721951708198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010881147533655166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015882111340761184, + "weighted_rkd_loss": 0.005440573766827583 + }, + { + "epoch": 6.314907872696818, + "flow_matching_loss": 0.01612699143588543, + "grad_norm": 0.17910315096378326, + "learning_rate": 4.702901515129989e-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": 0.004781160946004092, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018259460222907365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011388267949223518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01612699143588543, + "weighted_rkd_loss": 0.005694133974611759 + }, + { + "epoch": 6.331658291457287, + "flow_matching_loss": 0.01585171017795801, + "grad_norm": 0.1363685131072998, + "learning_rate": 4.630987517256734e-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": 0.004472851287573576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016937122913077475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010639414750039577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01585171017795801, + "weighted_rkd_loss": 0.005319707375019789 + }, + { + "epoch": 6.348408710217756, + "flow_matching_loss": 0.015469970647245646, + "grad_norm": 0.13690084218978882, + "learning_rate": 4.559527093580412e-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": 0.004476161859929562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019692603673320264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010921584069728851, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015469970647245646, + "weighted_rkd_loss": 0.0054607920348644255 + }, + { + "epoch": 6.365159128978225, + "flow_matching_loss": 0.017418408021330833, + "grad_norm": 0.1333802491426468, + "learning_rate": 4.488523370011456e-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": 0.004803127376362681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020921259187161923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011698380671441555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017418408021330833, + "weighted_rkd_loss": 0.005849190335720777 + }, + { + "epoch": 6.381909547738694, + "flow_matching_loss": 0.015745705273002387, + "grad_norm": 0.2102486789226532, + "learning_rate": 4.417979452482766e-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": 0.005005351454019547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002100074465852231, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012110777292400598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015745705273002387, + "weighted_rkd_loss": 0.006055388646200299 + }, + { + "epoch": 6.398659966499163, + "flow_matching_loss": 0.014893369190394879, + "grad_norm": 0.13255921006202698, + "learning_rate": 4.347898426813811e-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": 0.005195147078484297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002191038301680237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012581332447007298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014893369190394879, + "weighted_rkd_loss": 0.006290666223503649 + }, + { + "epoch": 6.415410385259632, + "flow_matching_loss": 0.015815038699656725, + "grad_norm": 0.1442941427230835, + "learning_rate": 4.278283358575694e-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": 0.004424917628057301, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00187108451500535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010720919771119952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015815038699656725, + "weighted_rkd_loss": 0.005360459885559976 + }, + { + "epoch": 6.432160804020101, + "flow_matching_loss": 0.016232730634510518, + "grad_norm": 0.1966170370578766, + "learning_rate": 4.2091372929570145e-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": 0.0037545856786891816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017783464048989118, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00928751789033413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016232730634510518, + "weighted_rkd_loss": 0.004643758945167065 + }, + { + "epoch": 6.44891122278057, + "flow_matching_loss": 0.0173671611584723, + "grad_norm": 0.16270123422145844, + "learning_rate": 4.1404632546306934e-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": 0.00426407353952527, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017715994967147709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010299746599048375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0173671611584723, + "weighted_rkd_loss": 0.005149873299524188 + }, + { + "epoch": 6.465661641541039, + "flow_matching_loss": 0.015224199555814267, + "grad_norm": 0.1519557237625122, + "learning_rate": 4.0722642476216286e-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": 0.004490367136895657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019021688785869628, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010882903169840575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015224199555814267, + "weighted_rkd_loss": 0.0054414515849202875 + }, + { + "epoch": 6.482412060301508, + "flow_matching_loss": 0.01661739433184266, + "grad_norm": 0.14995500445365906, + "learning_rate": 4.00454325517533e-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": 0.004987293970771134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022416375228203832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012216225545853377, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01661739433184266, + "weighted_rkd_loss": 0.0061081127729266885 + }, + { + "epoch": 6.499162479061977, + "flow_matching_loss": 0.015825352538377045, + "grad_norm": 0.16931673884391785, + "learning_rate": 3.937303239627393e-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": 0.004396289819851518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001814883027691394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010607462516054511, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015825352538377045, + "weighted_rkd_loss": 0.005303731258027256 + }, + { + "epoch": 6.515912897822446, + "flow_matching_loss": 0.01738193165510893, + "grad_norm": 0.14341159164905548, + "learning_rate": 3.870547142273918e-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": 0.0036826774710789325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013857228332199156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008751077763736247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01738193165510893, + "weighted_rkd_loss": 0.004375538881868124 + }, + { + "epoch": 6.532663316582915, + "flow_matching_loss": 0.01572351260110736, + "grad_norm": 0.16346746683120728, + "learning_rate": 3.8042778832428736e-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": 0.004531700373627245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017317167716100813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010795117449015379, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01572351260110736, + "weighted_rkd_loss": 0.005397558724507689 + }, + { + "epoch": 6.549413735343384, + "flow_matching_loss": 0.019746193196624517, + "grad_norm": 0.14955230057239532, + "learning_rate": 3.7384983613663195e-06, + "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": 0.004160749423317611, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018678586231544614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010189357539638877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019746193196624517, + "weighted_rkd_loss": 0.005094678769819438 + }, + { + "epoch": 6.566164154103853, + "flow_matching_loss": 0.018092065956443547, + "grad_norm": 0.17253366112709045, + "learning_rate": 3.6732114540536503e-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": 0.004915033443830907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020113187143579125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011841385671868921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018092065956443547, + "weighted_rkd_loss": 0.0059206928359344605 + }, + { + "epoch": 6.582914572864322, + "flow_matching_loss": 0.017865691613405944, + "grad_norm": 0.21846531331539154, + "learning_rate": 3.608420017165683e-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": 0.004521297919563949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002094288810621947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011136884754523634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017865691613405944, + "weighted_rkd_loss": 0.005568442377261817 + }, + { + "epoch": 6.599664991624791, + "flow_matching_loss": 0.016273113992065193, + "grad_norm": 0.14731267094612122, + "learning_rate": 3.5441268848897664e-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": 0.004396083881147206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017308910959400237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010523058846592903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016273113992065193, + "weighted_rkd_loss": 0.005261529423296452 + }, + { + "epoch": 6.61641541038526, + "flow_matching_loss": 0.016229044739156963, + "grad_norm": 0.13433827459812164, + "learning_rate": 3.480334869615792e-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": 0.003977289563044906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018249677086714655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009779546782374382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016229044739156963, + "weighted_rkd_loss": 0.004889773391187191 + }, + { + "epoch": 6.633165829145729, + "flow_matching_loss": 0.01615925971418619, + "grad_norm": 0.17934858798980713, + "learning_rate": 3.417046761813155e-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": 0.004120925953611731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002054018119815737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010295870155096054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01615925971418619, + "weighted_rkd_loss": 0.005147935077548027 + }, + { + "epoch": 6.649916247906198, + "flow_matching_loss": 0.01634274534881115, + "grad_norm": 0.15029595792293549, + "learning_rate": 3.3542653299087263e-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": 0.0038260640110820534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001958185143303126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009610313083976507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01634274534881115, + "weighted_rkd_loss": 0.004805156541988253 + }, + { + "epoch": 6.666666666666667, + "flow_matching_loss": 0.01320421127602458, + "grad_norm": 0.16808393597602844, + "learning_rate": 3.2919933201657117e-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": 0.004408310819417238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018108319956809283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010627453541383147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01320421127602458, + "weighted_rkd_loss": 0.0053137267706915734 + }, + { + "epoch": 6.683417085427136, + "flow_matching_loss": 0.01695033786818385, + "grad_norm": 0.175664022564888, + "learning_rate": 3.2302334565635584e-06, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004060668661259115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016366543364711105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009757991693913937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01695033786818385, + "weighted_rkd_loss": 0.004878995846956969 + }, + { + "epoch": 6.700167504187605, + "flow_matching_loss": 0.014768802467733622, + "grad_norm": 0.19610930979251862, + "learning_rate": 3.1689884406787674e-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": 0.0039969151839613914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018481418024748563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00984197212383151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014768802467733622, + "weighted_rkd_loss": 0.004920986061915755 + }, + { + "epoch": 6.716917922948074, + "flow_matching_loss": 0.01720321225002408, + "grad_norm": 0.16937272250652313, + "learning_rate": 3.1082609515667424e-06, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004821825376711786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019550187862478197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011598669551312923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01720321225002408, + "weighted_rkd_loss": 0.0057993347756564615 + }, + { + "epoch": 6.733668341708543, + "flow_matching_loss": 0.015799351502209902, + "grad_norm": 0.183858260512352, + "learning_rate": 3.048053645644594e-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": 0.004556529875844717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018371412064880134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010950201004743577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015799351502209902, + "weighted_rkd_loss": 0.005475100502371788 + }, + { + "epoch": 6.750418760469012, + "flow_matching_loss": 0.01648924229666591, + "grad_norm": 0.1715034395456314, + "learning_rate": 2.988369156574916e-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": 0.004184935172088444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016394643927924335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010009334795176983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01648924229666591, + "weighted_rkd_loss": 0.005004667397588492 + }, + { + "epoch": 6.767169179229481, + "flow_matching_loss": 0.016042891331017016, + "grad_norm": 0.13695967197418213, + "learning_rate": 2.9292100951506225e-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": 0.004686999879777431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018519218778237701, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01122592156752944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016042891331017016, + "weighted_rkd_loss": 0.00561296078376472 + }, + { + "epoch": 6.78391959798995, + "flow_matching_loss": 0.016733720898628235, + "grad_norm": 0.16754966974258423, + "learning_rate": 2.8705790491807033e-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": 0.0039164737798273565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017174503766000271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009550398029386998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016733720898628235, + "weighted_rkd_loss": 0.004775199014693499 + }, + { + "epoch": 6.800670016750419, + "flow_matching_loss": 0.014544985536485911, + "grad_norm": 0.15087449550628662, + "learning_rate": 2.8124785833770433e-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": 0.00489426520653069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020002575591206552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011788788065314294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014544985536485911, + "weighted_rkd_loss": 0.005894394032657147 + }, + { + "epoch": 6.817420435510888, + "flow_matching_loss": 0.016845161374658347, + "grad_norm": 0.18539758026599884, + "learning_rate": 2.754911239242241e-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": 0.004746794002130628, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021222085459157826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011615796573460102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016845161374658347, + "weighted_rkd_loss": 0.005807898286730051 + }, + { + "epoch": 6.834170854271357, + "flow_matching_loss": 0.01456131674349308, + "grad_norm": 0.17953617870807648, + "learning_rate": 2.6978795349584204e-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": 0.00482990201562643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018176681362092495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011477472074329854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01456131674349308, + "weighted_rkd_loss": 0.005738736037164927 + }, + { + "epoch": 6.850921273031826, + "flow_matching_loss": 0.016708610765635967, + "grad_norm": 0.1707935482263565, + "learning_rate": 2.641385965277088e-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": 0.004654878587462008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021587817347608505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0114685389213264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016708610765635967, + "weighted_rkd_loss": 0.0057342694606632 + }, + { + "epoch": 6.867671691792295, + "flow_matching_loss": 0.014694995246827602, + "grad_norm": 0.19747313857078552, + "learning_rate": 2.5854330014099923e-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": 0.0036429993459023537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015449580852873624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008830956788733602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014694995246827602, + "weighted_rkd_loss": 0.004415478394366801 + }, + { + "epoch": 6.884422110552764, + "flow_matching_loss": 0.01677362499758601, + "grad_norm": 0.12381528317928314, + "learning_rate": 2.5300230909210298e-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": 0.003627436398528516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018569727661088109, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009111845586448907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01677362499758601, + "weighted_rkd_loss": 0.004555922793224454 + }, + { + "epoch": 6.901172529313233, + "flow_matching_loss": 0.015464589465409517, + "grad_norm": 0.1643161177635193, + "learning_rate": 2.4751586576191983e-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": 0.00446525604929775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019164797675330191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010846991883590817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015464589465409517, + "weighted_rkd_loss": 0.0054234959417954086 + }, + { + "epoch": 6.917922948073702, + "flow_matching_loss": 0.015309206023812295, + "grad_norm": 0.1433928906917572, + "learning_rate": 2.420842101452538e-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": 0.004670693026855588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016811168286949396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011022503068670631, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015309206023812295, + "weighted_rkd_loss": 0.005511251534335316 + }, + { + "epoch": 6.934673366834171, + "flow_matching_loss": 0.015999872423708438, + "grad_norm": 0.158779114484787, + "learning_rate": 2.3670757984031816e-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": 0.004784934082999825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001908725016983226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011478593293577432, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015999872423708438, + "weighted_rkd_loss": 0.005739296646788716 + }, + { + "epoch": 6.95142378559464, + "flow_matching_loss": 0.0146644729655236, + "grad_norm": 0.14927701652050018, + "learning_rate": 2.313862100383404e-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": 0.003657519840635359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001735816220752895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009050855925306679, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0146644729655236, + "weighted_rkd_loss": 0.004525427962653339 + }, + { + "epoch": 6.968174204355109, + "flow_matching_loss": 0.01477276049554348, + "grad_norm": 0.18146266043186188, + "learning_rate": 2.2612033351327406e-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": 0.00446374686434865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017345455940812825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010662039229646324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01477276049554348, + "weighted_rkd_loss": 0.005331019614823162 + }, + { + "epoch": 6.984924623115578, + "flow_matching_loss": 0.01612620856612921, + "grad_norm": 0.1247067078948021, + "learning_rate": 2.2091018061161788e-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": 0.004328918852843344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019952588831074537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01065309657715261, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01612620856612921, + "weighted_rkd_loss": 0.005326548288576305 + }, + { + "epoch": 7.001675041876047, + "flow_matching_loss": 0.01582100810483098, + "grad_norm": 0.1982302963733673, + "learning_rate": 2.15755979242338e-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": 0.004648764641024173, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021950488444417715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011492578079923987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01582100810483098, + "weighted_rkd_loss": 0.005746289039961994 + }, + { + "epoch": 7.018425460636516, + "flow_matching_loss": 0.018236480932682753, + "grad_norm": 0.19694870710372925, + "learning_rate": 2.106579548669e-06, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005001381016336381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020459327613934876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012048694863915443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018236480932682753, + "weighted_rkd_loss": 0.0060243474319577215 + }, + { + "epoch": 7.035175879396985, + "flow_matching_loss": 0.014276114106178284, + "grad_norm": 0.1846606433391571, + "learning_rate": 2.0561633048940477e-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": 0.004038699180819094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016007982019800692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009678196534514428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014276114106178284, + "weighted_rkd_loss": 0.004839098267257214 + }, + { + "epoch": 7.051926298157454, + "flow_matching_loss": 0.016784136835485696, + "grad_norm": 0.15324676036834717, + "learning_rate": 2.006313266468355e-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": 0.004523446527309715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001994902826845646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011041795928031206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016784136835485696, + "weighted_rkd_loss": 0.005520897964015603 + }, + { + "epoch": 7.068676716917923, + "flow_matching_loss": 0.014370397850871085, + "grad_norm": 0.16374434530735016, + "learning_rate": 1.957031613994097e-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": 0.0038065645145252346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014738554309587926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009086984395980834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014370397850871085, + "weighted_rkd_loss": 0.004543492197990417 + }, + { + "epoch": 7.085427135678392, + "flow_matching_loss": 0.01604610849171877, + "grad_norm": 0.13959617912769318, + "learning_rate": 1.9083205032103995e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0047938420437276365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002127973339520395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011715657310560346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01604610849171877, + "weighted_rkd_loss": 0.005857828655280173 + }, + { + "epoch": 7.102177554438861, + "flow_matching_loss": 0.01643010778352618, + "grad_norm": 0.15431803464889526, + "learning_rate": 1.8601820648990536e-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": 0.003699658077675849, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016984055924694984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009097721613943576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01643010778352618, + "weighted_rkd_loss": 0.004548860806971788 + }, + { + "epoch": 7.11892797319933, + "flow_matching_loss": 0.014956025872379541, + "grad_norm": 0.15228129923343658, + "learning_rate": 1.8126184047912936e-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": 0.004501348035410047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00216407316038385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011166769126430155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014956025872379541, + "weighted_rkd_loss": 0.005583384563215077 + }, + { + "epoch": 7.135678391959799, + "flow_matching_loss": 0.015140857826918364, + "grad_norm": 0.1639655977487564, + "learning_rate": 1.7656316034757024e-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": 0.004590366035699844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018575215362943709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011038253549486399, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015140857826918364, + "weighted_rkd_loss": 0.005519126774743199 + }, + { + "epoch": 7.152428810720268, + "flow_matching_loss": 0.014944946672767401, + "grad_norm": 0.14010633528232574, + "learning_rate": 1.7192237163071816e-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": 0.004896424198523164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002066554082557559, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011859402479603886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014944946672767401, + "weighted_rkd_loss": 0.005929701239801943 + }, + { + "epoch": 7.169179229480737, + "flow_matching_loss": 0.016208091285079718, + "grad_norm": 0.16711357235908508, + "learning_rate": 1.6733967733170573e-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": 0.0037203690968453883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015936146199237555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009034352796152233, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016208091285079718, + "weighted_rkd_loss": 0.004517176398076117 + }, + { + "epoch": 7.185929648241206, + "flow_matching_loss": 0.014888655114918947, + "grad_norm": 0.15203069150447845, + "learning_rate": 1.628152779124279e-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": 0.004339498141780496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001999542448902503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010678538680076599, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014888655114918947, + "weighted_rkd_loss": 0.005339269340038299 + }, + { + "epoch": 7.202680067001675, + "flow_matching_loss": 0.014624554850161075, + "grad_norm": 0.11805262416601181, + "learning_rate": 1.583493712847715e-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": 0.0042874168837443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001616271329112351, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010191105119884015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014624554850161075, + "weighted_rkd_loss": 0.005095552559942007 + }, + { + "epoch": 7.219430485762144, + "flow_matching_loss": 0.017096712067723275, + "grad_norm": 0.13011588156223297, + "learning_rate": 1.5394215280196084e-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": 0.004473495739512145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017709834850393237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01071797488257289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017096712067723275, + "weighted_rkd_loss": 0.005358987441286445 + }, + { + "epoch": 7.236180904522613, + "flow_matching_loss": 0.015585697442293166, + "grad_norm": 0.18555836379528046, + "learning_rate": 1.4959381525000904e-06, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004154981835745275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016431983793154358, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009953161980956794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015585697442293166, + "weighted_rkd_loss": 0.004976580990478397 + }, + { + "epoch": 7.252931323283082, + "flow_matching_loss": 0.016454843897372486, + "grad_norm": 0.14526554942131042, + "learning_rate": 1.453045488392879e-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": 0.004315899196080863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001923052000347525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010554850520566105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016454843897372486, + "weighted_rkd_loss": 0.005277425260283053 + }, + { + "epoch": 7.269681742043551, + "flow_matching_loss": 0.015291227679699659, + "grad_norm": 0.12427052855491638, + "learning_rate": 1.4107454119620484e-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": 0.0038240417605265977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014376269245985895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00908571039326489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015291227679699659, + "weighted_rkd_loss": 0.004542855196632445 + }, + { + "epoch": 7.28643216080402, + "flow_matching_loss": 0.015831883577629925, + "grad_norm": 0.1394655555486679, + "learning_rate": 1.3690397735499778e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004891686118207872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002289877535076812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012073249882087111, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015831883577629925, + "weighted_rkd_loss": 0.0060366249410435556 + }, + { + "epoch": 7.303182579564489, + "flow_matching_loss": 0.01743771005421877, + "grad_norm": 0.14378996193408966, + "learning_rate": 1.327930397496403e-06, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005148415197618305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002286371891386807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012583202309906482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01743771005421877, + "weighted_rkd_loss": 0.006291601154953241 + }, + { + "epoch": 7.319932998324958, + "flow_matching_loss": 0.013519060239195824, + "grad_norm": 0.2160998433828354, + "learning_rate": 1.287419082058604e-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": 0.004035786679014564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019525676732882857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01002414091490209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013519060239195824, + "weighted_rkd_loss": 0.005012070457451045 + }, + { + "epoch": 7.336683417085427, + "flow_matching_loss": 0.015327712800353766, + "grad_norm": 0.1508062183856964, + "learning_rate": 1.2475075993327589e-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": 0.0042715171119198205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016422059037722648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010185240115970374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015327712800353766, + "weighted_rkd_loss": 0.005092620057985187 + }, + { + "epoch": 7.353433835845896, + "flow_matching_loss": 0.015343380253762006, + "grad_norm": 0.15916304290294647, + "learning_rate": 1.208197695176413e-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": 0.0046300238231197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002693808858748525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011953856470063329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015343380253762006, + "weighted_rkd_loss": 0.0059769282350316645 + }, + { + "epoch": 7.370184254606365, + "flow_matching_loss": 0.01437976472079754, + "grad_norm": 0.14108532667160034, + "learning_rate": 1.1694910891321259e-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": 0.004367586877197027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019202484982088208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010655422182753681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01437976472079754, + "weighted_rkd_loss": 0.005327711091376841 + }, + { + "epoch": 7.386934673366834, + "flow_matching_loss": 0.015470136143267154, + "grad_norm": 0.15935403108596802, + "learning_rate": 1.1313894743522363e-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": 0.003939059865660966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001673631521407515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009551751287654042, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015470136143267154, + "weighted_rkd_loss": 0.004775875643827021 + }, + { + "epoch": 7.403685092127303, + "flow_matching_loss": 0.016630494873970746, + "grad_norm": 0.20485074818134308, + "learning_rate": 1.0938945175248016e-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": 0.004128277907148004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001766240206779912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01002279594540596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016630494873970746, + "weighted_rkd_loss": 0.00501139797270298 + }, + { + "epoch": 7.420435510887772, + "flow_matching_loss": 0.01638146713376045, + "grad_norm": 0.10602868348360062, + "learning_rate": 1.0570078588007086e-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": 0.004238395742140711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017436613445170224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010220452863723039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01638146713376045, + "weighted_rkd_loss": 0.0051102264318615195 + }, + { + "epoch": 7.437185929648241, + "flow_matching_loss": 0.014488376211374998, + "grad_norm": 0.12349338829517365, + "learning_rate": 1.0207311117219009e-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": 0.0041768827708438035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017094228183850645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01006318824365735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014488376211374998, + "weighted_rkd_loss": 0.005031594121828675 + }, + { + "epoch": 7.45393634840871, + "flow_matching_loss": 0.014135436434298754, + "grad_norm": 0.11960218101739883, + "learning_rate": 9.850658631508198e-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": 0.003858961584046483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016201299498789013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009338053129613399, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014135436434298754, + "weighted_rkd_loss": 0.004669026564806699 + }, + { + "epoch": 7.4706867671691795, + "flow_matching_loss": 0.015964719746261836, + "grad_norm": 0.15216432511806488, + "learning_rate": 9.500136732009751e-07, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038564404007047416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00145489246933721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009167773276567459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015964719746261836, + "weighted_rkd_loss": 0.004583886638283729 + }, + { + "epoch": 7.4874371859296485, + "flow_matching_loss": 0.013642612006515264, + "grad_norm": 0.14385931193828583, + "learning_rate": 9.155760751687054e-07, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039043609518557786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017102246696595103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009518946800380946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013642612006515264, + "weighted_rkd_loss": 0.004759473400190473 + }, + { + "epoch": 7.5041876046901175, + "flow_matching_loss": 0.014753709733486175, + "grad_norm": 0.16965070366859436, + "learning_rate": 8.817545754661155e-07, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00408722497522831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001713181775994599, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009887631935998797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014753709733486175, + "weighted_rkd_loss": 0.004943815967999398 + }, + { + "epoch": 7.5209380234505865, + "flow_matching_loss": 0.016744329128414392, + "grad_norm": 0.16780605912208557, + "learning_rate": 8.485506535551668e-07, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004479729966260493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019441045355051757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01090356446802616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016744329128414392, + "weighted_rkd_loss": 0.00545178223401308 + }, + { + "epoch": 7.5376884422110555, + "flow_matching_loss": 0.014030791632831097, + "grad_norm": 0.1584310531616211, + "learning_rate": 8.159657618829685e-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": 0.003778664767742157, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016007275553420186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009158057114109396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014030791632831097, + "weighted_rkd_loss": 0.004579028557054698 + }, + { + "epoch": 7.5544388609715245, + "flow_matching_loss": 0.016403439175337552, + "grad_norm": 0.14649353921413422, + "learning_rate": 7.840013258182471e-07, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003972286405041814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014566533092875034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009401226090267301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016403439175337552, + "weighted_rkd_loss": 0.0047006130451336505 + }, + { + "epoch": 7.5711892797319935, + "flow_matching_loss": 0.016914738900959493, + "grad_norm": 0.12157639861106873, + "learning_rate": 7.526587435889809e-07, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003476941166445613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013800669636111706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008333949279040099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016914738900959493, + "weighted_rkd_loss": 0.004166974639520049 + }, + { + "epoch": 7.5879396984924625, + "flow_matching_loss": 0.01702210819348693, + "grad_norm": 0.16324695944786072, + "learning_rate": 7.219393862212593e-07, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004177236021496356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001766754686832428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010121226636692881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01702210819348693, + "weighted_rkd_loss": 0.005060613318346441 + }, + { + "epoch": 7.6046901172529315, + "flow_matching_loss": 0.016736098565161228, + "grad_norm": 0.13862305879592896, + "learning_rate": 6.918445974792819e-07, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035174431046471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017616379889659583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00879652420990169, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016736098565161228, + "weighted_rkd_loss": 0.004398262104950845 + }, + { + "epoch": 7.6214405360134005, + "flow_matching_loss": 0.015986019000411033, + "grad_norm": 0.1334938406944275, + "learning_rate": 6.6237569380661e-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": 0.003616045939270407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001374229503562674, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008606321271508932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015986019000411033, + "weighted_rkd_loss": 0.004303160635754466 + }, + { + "epoch": 7.63819095477387, + "flow_matching_loss": 0.01590715404599905, + "grad_norm": 0.14786288142204285, + "learning_rate": 6.335339642685495e-07, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004220075206831098, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017576340294908732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01019778442569077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01590715404599905, + "weighted_rkd_loss": 0.005098892212845385 + }, + { + "epoch": 7.654941373534339, + "flow_matching_loss": 0.015476820152252912, + "grad_norm": 0.1825518161058426, + "learning_rate": 6.05320670495787e-07, + "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": 0.0040603889618068935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017880713276099414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009908849187195302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015476820152252912, + "weighted_rkd_loss": 0.004954424593597651 + }, + { + "epoch": 7.671691792294808, + "flow_matching_loss": 0.014729360304772853, + "grad_norm": 0.18547452986240387, + "learning_rate": 5.777370466291954e-07, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004903944488614797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002282101463060826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012089990358799697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014729360304772853, + "weighted_rkd_loss": 0.006044995179399848 + }, + { + "epoch": 7.688442211055277, + "flow_matching_loss": 0.01499204495921731, + "grad_norm": 0.17845290899276733, + "learning_rate": 5.507842992658336e-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": 0.003743468993343413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015168702288065107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00900380820967257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01499204495921731, + "weighted_rkd_loss": 0.004501904104836285 + }, + { + "epoch": 7.705192629815746, + "flow_matching_loss": 0.01654659826308489, + "grad_norm": 0.19419962167739868, + "learning_rate": 5.24463607406192e-07, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004446967015974223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017932968446984887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010687231039628386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01654659826308489, + "weighted_rkd_loss": 0.005343615519814193 + }, + { + "epoch": 7.721943048576215, + "flow_matching_loss": 0.014629477728158235, + "grad_norm": 0.16444142162799835, + "learning_rate": 4.987761224025944e-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": 0.004281575768254697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017087026499211788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010271854186430573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014629477728158235, + "weighted_rkd_loss": 0.005135927093215286 + }, + { + "epoch": 7.738693467336684, + "flow_matching_loss": 0.01777318762615323, + "grad_norm": 0.13727985322475433, + "learning_rate": 4.737229679088556e-07, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036333360243588687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016103479429148138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008877020049840211, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01777318762615323, + "weighted_rkd_loss": 0.004438510024920106 + }, + { + "epoch": 7.755443886097153, + "flow_matching_loss": 0.016660143435001374, + "grad_norm": 0.13840682804584503, + "learning_rate": 4.493052398311115e-07, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004052148922346533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018098844622727484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009914182359352708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016660143435001374, + "weighted_rkd_loss": 0.004957091179676354 + }, + { + "epoch": 7.772194304857622, + "flow_matching_loss": 0.015904311370104552, + "grad_norm": 0.18793518841266632, + "learning_rate": 4.2552400627989043e-07, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004106870060786605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017587932874448598, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009972533490508795, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015904311370104552, + "weighted_rkd_loss": 0.004986266745254398 + }, + { + "epoch": 7.788944723618091, + "flow_matching_loss": 0.015392695367336274, + "grad_norm": 0.22930394113063812, + "learning_rate": 4.023803075233923e-07, + "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": 0.004883068730123341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002066736004780978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011832873662933707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015392695367336274, + "weighted_rkd_loss": 0.0059164368314668534 + }, + { + "epoch": 7.80569514237856, + "flow_matching_loss": 0.01676181796938181, + "grad_norm": 0.1597457081079483, + "learning_rate": 3.7987515594196997e-07, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036124582402408123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001387284556403756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00861220108345151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01676181796938181, + "weighted_rkd_loss": 0.004306100541725755 + }, + { + "epoch": 7.822445561139029, + "flow_matching_loss": 0.015276381466537713, + "grad_norm": 0.22032739222049713, + "learning_rate": 3.580095359838648e-07, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004534202092327177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016930141719058156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010761418426409364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015276381466537713, + "weighted_rkd_loss": 0.005380709213204682 + }, + { + "epoch": 7.839195979899498, + "flow_matching_loss": 0.015883130393922328, + "grad_norm": 0.19140352308750153, + "learning_rate": 3.3678440412212576e-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": 0.003926279791630805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016084841859992593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009461043868213892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015883130393922328, + "weighted_rkd_loss": 0.004730521934106946 + }, + { + "epoch": 7.855946398659967, + "flow_matching_loss": 0.016306853387504816, + "grad_norm": 0.12501484155654907, + "learning_rate": 3.162006888127833e-07, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004368531377986073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019060501595959066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010643112845718861, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016306853387504816, + "weighted_rkd_loss": 0.0053215564228594305 + }, + { + "epoch": 7.872696817420436, + "flow_matching_loss": 0.017385248746722936, + "grad_norm": 0.1701114922761917, + "learning_rate": 2.9625929045422327e-07, + "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": 0.00435904401820153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001872993551660329, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010591081576421857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017385248746722936, + "weighted_rkd_loss": 0.0052955407882109284 + }, + { + "epoch": 7.889447236180905, + "flow_matching_loss": 0.017262033838778734, + "grad_norm": 0.1395217478275299, + "learning_rate": 2.7696108134781186e-07, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036485470365732907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013480199966579675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008645114209502935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017262033838778734, + "weighted_rkd_loss": 0.004322557104751467 + }, + { + "epoch": 7.906197654941374, + "flow_matching_loss": 0.0179462818428874, + "grad_norm": 0.1584036499261856, + "learning_rate": 2.583069056597376e-07, + "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": 0.004363359813578427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018044174241367728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010531136952340602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0179462818428874, + "weighted_rkd_loss": 0.005265568476170301 + }, + { + "epoch": 7.922948073701843, + "flow_matching_loss": 0.014670177130028605, + "grad_norm": 0.16058886051177979, + "learning_rate": 2.4029757938407116e-07, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004476101510226727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019054161792155355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010857619298622011, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014670177130028605, + "weighted_rkd_loss": 0.005428809649311006 + }, + { + "epoch": 7.939698492462312, + "flow_matching_loss": 0.017554907780140638, + "grad_norm": 0.12912341952323914, + "learning_rate": 2.229338903070871e-07, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00388623287435621, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001478163164574653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009250628901645542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017554907780140638, + "weighted_rkd_loss": 0.004625314450822771 + }, + { + "epoch": 7.956448911222781, + "flow_matching_loss": 0.016205093823373318, + "grad_norm": 0.12646424770355225, + "learning_rate": 2.0621659797279645e-07, + "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": 0.0042540517169982195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017718847666401416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01027998821809888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016205093823373318, + "weighted_rkd_loss": 0.00513999410904944 + }, + { + "epoch": 7.97319932998325, + "flow_matching_loss": 0.016403627768158912, + "grad_norm": 0.18282198905944824, + "learning_rate": 1.901464336497233e-07, + "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": 0.004610209143720567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019326368696056307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011153055075556039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016403627768158912, + "weighted_rkd_loss": 0.0055765275377780196 + }, + { + "epoch": 7.989949748743719, + "flow_matching_loss": 0.015249314811080694, + "grad_norm": 0.17364180088043213, + "learning_rate": 1.7472410029891527e-07, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00389033502433449, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001546752726426348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009327422780916094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015249314811080694, + "weighted_rkd_loss": 0.004663711390458047 + }, + { + "epoch": 8.006700167504187, + "flow_matching_loss": 0.014920513145625591, + "grad_norm": 0.19061186909675598, + "learning_rate": 1.5995027254319173e-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": 0.0035726051311939953, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014776690921280533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008622879348695277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014920513145625591, + "weighted_rkd_loss": 0.004311439674347639 + }, + { + "epoch": 8.023450586264657, + "flow_matching_loss": 0.016243462171405553, + "grad_norm": 0.12961623072624207, + "learning_rate": 1.4582559663764372e-07, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038010131334885954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001989023806527257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00959105002693832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016243462171405553, + "weighted_rkd_loss": 0.00479552501346916 + }, + { + "epoch": 8.040201005025125, + "flow_matching_loss": 0.017470394633710383, + "grad_norm": 0.17603987455368042, + "learning_rate": 1.323506904413485e-07, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038247731747105716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016475262236781418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009297072421759367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017470394633710383, + "weighted_rkd_loss": 0.0046485362108796835 + }, + { + "epoch": 8.056951423785595, + "flow_matching_loss": 0.0145564878359437, + "grad_norm": 0.197255939245224, + "learning_rate": 1.195261433903594e-07, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042893028585240245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017500674701295793, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010328673152253031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0145564878359437, + "weighted_rkd_loss": 0.005164336576126516 + }, + { + "epoch": 8.073701842546063, + "flow_matching_loss": 0.014916270691901445, + "grad_norm": 0.1560496985912323, + "learning_rate": 1.0735251647190969e-07, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003976433514617383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015865480876527728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009539415081962942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014916270691901445, + "weighted_rkd_loss": 0.004769707540981471 + }, + { + "epoch": 8.090452261306533, + "flow_matching_loss": 0.014834001008421183, + "grad_norm": 0.1427733153104782, + "learning_rate": 9.583034219987408e-08, + "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": 0.003930709883570671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015655665658414364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009426986379548908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014834001008421183, + "weighted_rkd_loss": 0.004713493189774454 + }, + { + "epoch": 8.107202680067001, + "flow_matching_loss": 0.013917479477822781, + "grad_norm": 0.1223326325416565, + "learning_rate": 8.496012459148717e-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": 0.0044130357215181, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002009857736993581, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010835929354652763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013917479477822781, + "weighted_rkd_loss": 0.0054179646773263816 + }, + { + "epoch": 8.12395309882747, + "flow_matching_loss": 0.017225508112460375, + "grad_norm": 0.11976644396781921, + "learning_rate": 7.474233914527618e-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": 0.003419135557487607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013502314046490937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00818850235082209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017225508112460375, + "weighted_rkd_loss": 0.004094251175411045 + }, + { + "epoch": 8.140703517587939, + "flow_matching_loss": 0.018462932761758567, + "grad_norm": 0.11892782896757126, + "learning_rate": 6.517743282028254e-08, + "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": 0.004395476914942265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017995877482462675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010590541595593094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018462932761758567, + "weighted_rkd_loss": 0.005295270797796547 + }, + { + "epoch": 8.157453936348409, + "flow_matching_loss": 0.01574688982218504, + "grad_norm": 0.13970652222633362, + "learning_rate": 5.6265824016492604e-08, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004414934199303389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001722412754315883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010552281187847257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01574688982218504, + "weighted_rkd_loss": 0.0052761405939236285 + }, + { + "epoch": 8.174204355108877, + "flow_matching_loss": 0.01800237987190485, + "grad_norm": 0.12215915322303772, + "learning_rate": 4.800790255654719e-08, + "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": 0.003544078720733523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001673859596485272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00876201707869768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01800237987190485, + "weighted_rkd_loss": 0.00438100853934884 + }, + { + "epoch": 8.190954773869347, + "flow_matching_loss": 0.015847709495574236, + "grad_norm": 0.15015308558940887, + "learning_rate": 4.040402966867862e-08, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004173338948749006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015449339873157442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009891611896455288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015847709495574236, + "weighted_rkd_loss": 0.004945805948227644 + }, + { + "epoch": 8.207705192629815, + "flow_matching_loss": 0.015503469109535217, + "grad_norm": 0.20167481899261475, + "learning_rate": 3.3454537970925014e-08, + "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": 0.004341295640915632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021772518870420755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010859843250364066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015503469109535217, + "weighted_rkd_loss": 0.005429921625182033 + }, + { + "epoch": 8.224455611390285, + "flow_matching_loss": 0.01842154674232006, + "grad_norm": 0.19743748009204865, + "learning_rate": 2.715973145656192e-08, + "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": 0.0035309243947267534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015276472142431885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0085894959513098, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01842154674232006, + "weighted_rkd_loss": 0.0042947479756549 + }, + { + "epoch": 8.241206030150753, + "flow_matching_loss": 0.014870964922010899, + "grad_norm": 0.11341430246829987, + "learning_rate": 2.151988548082129e-08, + "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": 0.003632827172987163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014750438393093645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008740698220208287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014870964922010899, + "weighted_rkd_loss": 0.004370349110104144 + }, + { + "epoch": 8.257956448911223, + "flow_matching_loss": 0.015233697462826967, + "grad_norm": 0.18681037425994873, + "learning_rate": 1.6535246748837772e-08, + "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": 0.004151923372410238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019706888357177376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01027453546412289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015233697462826967, + "weighted_rkd_loss": 0.005137267732061445 + }, + { + "epoch": 8.274706867671691, + "flow_matching_loss": 0.017054551746696234, + "grad_norm": 0.15061014890670776, + "learning_rate": 1.2206033304855702e-08, + "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": 0.0037409895565360786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016018183785490691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00908379745669663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017054551746696234, + "weighted_rkd_loss": 0.004541898728348315 + }, + { + "epoch": 8.291457286432161, + "flow_matching_loss": 0.014949669688940048, + "grad_norm": 0.13207487761974335, + "learning_rate": 8.532434522695031e-09, + "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": 0.004302018298767507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017844451940618455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010388481616973876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014949669688940048, + "weighted_rkd_loss": 0.005194240808486938 + }, + { + "epoch": 8.308207705192629, + "flow_matching_loss": 0.01610209411010146, + "grad_norm": 0.14128775894641876, + "learning_rate": 5.5146110974679805e-09, + "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": 0.0036398351658135653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001506704674102366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00878637502901256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01610209411010146, + "weighted_rkd_loss": 0.00439318751450628 + }, + { + "epoch": 8.324958123953099, + "flow_matching_loss": 0.016640785429626704, + "grad_norm": 0.14564873278141022, + "learning_rate": 3.1526950385496467e-09, + "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": 0.003576566115953028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014642269699834288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00861735912039876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016640785429626704, + "weighted_rkd_loss": 0.00430867956019938 + }, + { + "epoch": 8.341708542713567, + "flow_matching_loss": 0.013791149761527777, + "grad_norm": 0.139508917927742, + "learning_rate": 1.4467896637959666e-09, + "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": 0.0038056798744946717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014783959602937102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009089755592867732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013791149761527777, + "weighted_rkd_loss": 0.004544877796433866 + }, + { + "epoch": 8.358458961474037, + "flow_matching_loss": 0.015681551396846773, + "grad_norm": 0.11605437844991684, + "learning_rate": 3.9696959503898733e-10, + "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": 0.004223085520789027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017177864559926093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010163957532495261, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015681551396846773, + "weighted_rkd_loss": 0.0050819787662476305 + }, + { + "epoch": 8.375209380234505, + "flow_matching_loss": 0.014806061424314976, + "grad_norm": 0.19967901706695557, + "learning_rate": 3.280754806156416e-12, + "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": 0.004259938187897205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020176144782453774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010537490900605916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014806061424314976, + "weighted_rkd_loss": 0.005268745450302958 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": true + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/config.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..27700cac5d292d4d42897b3c2e80bdc58e55c60e --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.5719731235192254, + -2.1489371002267235, + 0.7003612741268546 + ], + "std": [ + 1.669945884010784, + 1.3739411381361726, + 0.0009413282367754987 + ], + "q01": [ + -0.7944286339399775, + -5.486611518471242, + 0.699977555971739 + ], + "q99": [ + 6.019886905263648, + -0.5229184210026855, + 0.702930326282967 + ] + }, + "base_rotation": { + "max": [ + 0.0, + 0.0, + 1.0, + 1.0 + ], + "min": [ + 0.0, + 0.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0, + 0.0, + 0.248543484335273, + 0.5959317405459531 + ], + "std": [ + 0.0, + 0.0, + 0.6655550240074092, + 0.374283995143559 + ], + "q01": [ + 0.0, + 0.0, + -1.0000000000000002, + 4.503254968910557e-06 + ], + "q99": [ + 0.0, + 0.0, + 1.0000000000000002, + 1.0000000000000002 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.2465490874130862, + -0.004711658908481467, + 0.43463835464312073 + ], + "std": [ + 0.15798784670671595, + 0.25624102400497073, + 0.20836028727974915 + ], + "q01": [ + -0.020752051134640608, + -0.4837933887347875, + 0.1802942952289915 + ], + "q99": [ + 0.5386784714844007, + 0.4801424212024721, + 0.695095732975766 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22531859714056926, + 0.046973702821026614, + -0.0568176967878771, + 0.12364307042955697 + ], + "std": [ + 0.8502144333612052, + 0.29591250318147405, + 0.30218053156837016, + 0.1636031353412931 + ], + "q01": [ + -0.9959057387172566, + -0.5665384745573048, + -0.5329334620834977, + 0.0014372279999397553 + ], + "q99": [ + 0.9904107951565839, + 0.6359623612021396, + 0.5026076022439377, + 0.42955559234879914 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03188757060429838, + -0.0321793333086053 + ], + "std": [ + 0.01171052313737148, + 0.011427336801203189 + ], + "q01": [ + 0.006462901319459925, + -0.040557140599419166 + ], + "q99": [ + 0.04037619335388386, + -0.006523947835993347 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006639007355857881, + 0.0027330421795561874, + -0.0017774441970214892, + 0.0 + ], + "std": [ + 0.11841528114040736, + 0.1041359083946544, + 0.08346265521208351, + 0.0 + ], + "q01": [ + -0.17562359675270237, + -0.21706402321529558, + -0.16273392585723104, + 0.0 + ], + "q99": [ + 0.2572918701644887, + 0.27022325575787115, + 0.1369626751049278, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9205539608927732 + ], + "std": [ + 0.39063657242534666 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.17711863866562935 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.00924803964621179, + 0.004053841761390575, + -0.08385115578460103 + ], + "std": [ + 0.45413072518223824, + 0.40158612714634856, + 0.41068634292849304 + ], + "q01": [ + -0.9669817145875343, + -0.9241307799783718, + -0.8264079503068825 + ], + "q99": [ + 0.8021511975722166, + 0.9241346582748478, + 0.8348116781502298 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.007644157504353809, + -0.019962543093886335, + -0.002047110452602617 + ], + "std": [ + 0.11076012154791087, + 0.13176735140285795, + 0.12286528665895614 + ], + "q01": [ + -0.2659105439401192, + -0.3961694033813533, + -0.3434870832468936 + ], + "q99": [ + 0.3145786959266369, + 0.31309430361227547, + 0.34107754401942986 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.29870612267533575 + ], + "std": [ + 0.9543487555375691 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7610674802484976 + ] + } + } + }, + "modalities": { + "video": { + "robot0_eye_in_hand": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_left": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_right": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + } + }, + "state": { + "base_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "base_rotation": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "end_effector_position_relative": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation_relative": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "gripper_qpos": { + "absolute": true, + "rotation_type": null, + "shape": [ + 2 + ], + "continuous": true + } + }, + "action": { + "base_motion": { + "absolute": true, + "rotation_type": null, + "shape": [ + 4 + ], + "continuous": true + }, + "control_mode": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + }, + "end_effector_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation": { + "absolute": true, + "rotation_type": "axis_angle", + "shape": [ + 3 + ], + "continuous": true + }, + "gripper_close": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + } + } + }, + "embodiment_tag": "new_embodiment" + } +} \ No newline at end of file diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a3758b2dac9cab1274647a0a3e4647bb197ec7fc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6f702e66e76c9c2a5773af5ec6b03a2aabb911ab03f04ce0b6ef1d2a8dca66a +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b23085757fd91fae834f4c6403f9a84dfd99ffc0 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6242297ba3d089711e642949f60242f06dbd7988bca7c9b6335962c0a8817a25 +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_5k/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..2249ec14e3a113c8a4f1e29a6206671aec6d02e9 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed1/default/train_5shot_seed1_processingline_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": 0.20356450378894805, + "grad_norm": 1.0490323305130005, + "learning_rate": 1.08e-06, + "loss": 0.4004, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1398262470960617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11395216882228851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.39360466301441194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.20356450378894805, + "weighted_rkd_loss": 0.19680233150720597 + }, + { + "epoch": 0.03350083752093802, + "flow_matching_loss": 0.21135056614875794, + "grad_norm": 0.8279632925987244, + "learning_rate": 2.28e-06, + "loss": 0.3916, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12995845153927804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10064354985952377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3605604529380798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.21135056614875794, + "weighted_rkd_loss": 0.1802802264690399 + }, + { + "epoch": 0.05025125628140704, + "flow_matching_loss": 0.19959198087453842, + "grad_norm": 0.8481504321098328, + "learning_rate": 3.48e-06, + "loss": 0.3868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13302332013845444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1083116926252842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.374358332157135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.19959198087453842, + "weighted_rkd_loss": 0.1871791660785675 + }, + { + "epoch": 0.06700167504187604, + "flow_matching_loss": 0.21401427686214447, + "grad_norm": 1.1905933618545532, + "learning_rate": 4.68e-06, + "loss": 0.3892, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12562137991189956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09915168732404708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3503944456577301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.21401427686214447, + "weighted_rkd_loss": 0.17519722282886505 + }, + { + "epoch": 0.08375209380234507, + "flow_matching_loss": 0.18060285598039627, + "grad_norm": 0.9187169075012207, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11310266405344009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08473661690950393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3109419494867325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.18060285598039627, + "weighted_rkd_loss": 0.15547097474336624 + }, + { + "epoch": 0.10050251256281408, + "flow_matching_loss": 0.1746109426021576, + "grad_norm": 0.9370708465576172, + "learning_rate": 7.08e-06, + "loss": 0.3298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11362355500459671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08309690579771996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.31034401059150696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.1746109426021576, + "weighted_rkd_loss": 0.15517200529575348 + }, + { + "epoch": 0.11725293132328309, + "flow_matching_loss": 0.16165669709444047, + "grad_norm": 0.7368708848953247, + "learning_rate": 8.28e-06, + "loss": 0.3148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11314839944243431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07991818860173225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30621498823165894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.16165669709444047, + "weighted_rkd_loss": 0.15310749411582947 + }, + { + "epoch": 0.13400335008375208, + "flow_matching_loss": 0.15373115316033364, + "grad_norm": 0.8529441356658936, + "learning_rate": 9.48e-06, + "loss": 0.2947, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10446386188268661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0729650903493166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.28189281225204466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.15373115316033364, + "weighted_rkd_loss": 0.14094640612602233 + }, + { + "epoch": 0.1507537688442211, + "flow_matching_loss": 0.12661252617836, + "grad_norm": 0.7936683893203735, + "learning_rate": 1.068e-05, + "loss": 0.2582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09823484048247337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06663991399109363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2631095930933952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.12661252617836, + "weighted_rkd_loss": 0.1315547965466976 + }, + { + "epoch": 0.16750418760469013, + "flow_matching_loss": 0.10844202637672425, + "grad_norm": 0.6985727548599243, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.2325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09290647432208061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.062390904128551486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24820385426282882, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.10844202637672425, + "weighted_rkd_loss": 0.12410192713141441 + }, + { + "epoch": 0.18425460636515914, + "flow_matching_loss": 0.09378867372870445, + "grad_norm": 0.6144883036613464, + "learning_rate": 1.308e-05, + "loss": 0.2126, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08944455981254577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05871000774204731, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23759912848472595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.09378867372870445, + "weighted_rkd_loss": 0.11879956424236297 + }, + { + "epoch": 0.20100502512562815, + "flow_matching_loss": 0.08636006340384483, + "grad_norm": 0.519714891910553, + "learning_rate": 1.428e-05, + "loss": 0.1946, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08254799991846085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0513349212706089, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21643092334270478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08636006340384483, + "weighted_rkd_loss": 0.10821546167135239 + }, + { + "epoch": 0.21775544388609716, + "flow_matching_loss": 0.07800828218460083, + "grad_norm": 0.6437940001487732, + "learning_rate": 1.548e-05, + "loss": 0.1787, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07631000094115734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.048831694200634955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20145169496536255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07800828218460083, + "weighted_rkd_loss": 0.10072584748268128 + }, + { + "epoch": 0.23450586264656617, + "flow_matching_loss": 0.08341701701283455, + "grad_norm": 0.6835018992424011, + "learning_rate": 1.6680000000000003e-05, + "loss": 0.1816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0760519951581955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04422654993832111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19633053839206696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.08341701701283455, + "weighted_rkd_loss": 0.09816526919603348 + }, + { + "epoch": 0.25125628140703515, + "flow_matching_loss": 0.07501983791589736, + "grad_norm": 0.6060196161270142, + "learning_rate": 1.7879999999999998e-05, + "loss": 0.1644, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06893355883657933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04092012867331505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17878724485635758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07501983791589736, + "weighted_rkd_loss": 0.08939362242817879 + }, + { + "epoch": 0.26800670016750416, + "flow_matching_loss": 0.07575965747237205, + "grad_norm": 0.5369148254394531, + "learning_rate": 1.908e-05, + "loss": 0.1611, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06494721546769142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.040794138610363004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17068857252597808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07575965747237205, + "weighted_rkd_loss": 0.08534428626298904 + }, + { + "epoch": 0.2847571189279732, + "flow_matching_loss": 0.07186435237526893, + "grad_norm": 0.4956815242767334, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06285238228738307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03799100033938885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1636957660317421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07186435237526893, + "weighted_rkd_loss": 0.08184788301587105 + }, + { + "epoch": 0.3015075376884422, + "flow_matching_loss": 0.06954732127487659, + "grad_norm": 0.4060089886188507, + "learning_rate": 2.148e-05, + "loss": 0.1457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05899851806461811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034271294809877874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1522683322429657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06954732127487659, + "weighted_rkd_loss": 0.07613416612148285 + }, + { + "epoch": 0.31825795644891125, + "flow_matching_loss": 0.07116162888705731, + "grad_norm": 0.4704577922821045, + "learning_rate": 2.268e-05, + "loss": 0.1464, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05825379975140095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033995861932635305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15050345957279204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07116162888705731, + "weighted_rkd_loss": 0.07525172978639602 + }, + { + "epoch": 0.33500837520938026, + "flow_matching_loss": 0.07152839414775372, + "grad_norm": 0.4596906900405884, + "learning_rate": 2.3880000000000002e-05, + "loss": 0.1431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05577990598976612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03162169083952904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14318150281906128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.07152839414775372, + "weighted_rkd_loss": 0.07159075140953064 + }, + { + "epoch": 0.35175879396984927, + "flow_matching_loss": 0.06675980575382709, + "grad_norm": 0.46783247590065, + "learning_rate": 2.508e-05, + "loss": 0.1351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.052671773731708525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0314154077321291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13675895631313323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06675980575382709, + "weighted_rkd_loss": 0.06837947815656661 + }, + { + "epoch": 0.3685092127303183, + "flow_matching_loss": 0.06015884168446064, + "grad_norm": 0.4413231313228607, + "learning_rate": 2.628e-05, + "loss": 0.12, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04683331362903118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02598679680377245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11965342536568642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06015884168446064, + "weighted_rkd_loss": 0.05982671268284321 + }, + { + "epoch": 0.3852596314907873, + "flow_matching_loss": 0.05922078154981136, + "grad_norm": 0.4952040910720825, + "learning_rate": 2.748e-05, + "loss": 0.1187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04629960469901562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02644723802804947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1190464474260807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05922078154981136, + "weighted_rkd_loss": 0.05952322371304035 + }, + { + "epoch": 0.4020100502512563, + "flow_matching_loss": 0.06449385173618793, + "grad_norm": 0.48206594586372375, + "learning_rate": 2.868e-05, + "loss": 0.1224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04490597955882549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025913392566144465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11572535187005997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.06449385173618793, + "weighted_rkd_loss": 0.057862675935029986 + }, + { + "epoch": 0.4187604690117253, + "flow_matching_loss": 0.054776575043797494, + "grad_norm": 0.40395382046699524, + "learning_rate": 2.9880000000000002e-05, + "loss": 0.1104, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04352367855608463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024162129126489164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11120948567986488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.054776575043797494, + "weighted_rkd_loss": 0.05560474283993244 + }, + { + "epoch": 0.4355108877721943, + "flow_matching_loss": 0.05850586108863354, + "grad_norm": 0.4056771993637085, + "learning_rate": 2.9999734259635676e-05, + "loss": 0.1118, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041912122443318364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022748018987476824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10657226219773293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05850586108863354, + "weighted_rkd_loss": 0.053286131098866464 + }, + { + "epoch": 0.45226130653266333, + "flow_matching_loss": 0.062202782928943635, + "grad_norm": 0.4282694160938263, + "learning_rate": 2.9998815663057245e-05, + "loss": 0.1095, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036747928336262704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02105743968859315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0945532962679863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.062202782928943635, + "weighted_rkd_loss": 0.04727664813399315 + }, + { + "epoch": 0.46901172529313234, + "flow_matching_loss": 0.051905983313918116, + "grad_norm": 0.39980289340019226, + "learning_rate": 2.999724096969228e-05, + "loss": 0.0978, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03628676719963551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019146752916276454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09172028750181198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.051905983313918116, + "weighted_rkd_loss": 0.04586014375090599 + }, + { + "epoch": 0.48576214405360135, + "flow_matching_loss": 0.05202174261212349, + "grad_norm": 0.42642244696617126, + "learning_rate": 2.9995010248422984e-05, + "loss": 0.0982, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03656212091445923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01933057326823473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0924548164010048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05202174261212349, + "weighted_rkd_loss": 0.0462274082005024 + }, + { + "epoch": 0.5025125628140703, + "flow_matching_loss": 0.048975753411650655, + "grad_norm": 0.3835107982158661, + "learning_rate": 2.999212359682832e-05, + "loss": 0.0927, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034414700232446194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018641455005854367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08747085481882096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.048975753411650655, + "weighted_rkd_loss": 0.04373542740941048 + }, + { + "epoch": 0.5192629815745393, + "flow_matching_loss": 0.05028046742081642, + "grad_norm": 0.3581988513469696, + "learning_rate": 2.9988581141179787e-05, + "loss": 0.0919, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03313282225281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016972948610782624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08323859423398972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05028046742081642, + "weighted_rkd_loss": 0.04161929711699486 + }, + { + "epoch": 0.5360134003350083, + "flow_matching_loss": 0.048362468183040616, + "grad_norm": 0.3743322193622589, + "learning_rate": 2.998438303643587e-05, + "loss": 0.0896, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03231747522950172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017821643501520157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0824565939605236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.048362468183040616, + "weighted_rkd_loss": 0.0412282969802618 + }, + { + "epoch": 0.5527638190954773, + "flow_matching_loss": 0.04962967894971371, + "grad_norm": 0.37111181020736694, + "learning_rate": 2.997952946623527e-05, + "loss": 0.0898, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031945614889264104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016547599248588084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08043882995843887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04962967894971371, + "weighted_rkd_loss": 0.04021941497921944 + }, + { + "epoch": 0.5695142378559463, + "flow_matching_loss": 0.050852949172258376, + "grad_norm": 0.418589323759079, + "learning_rate": 2.9974020642888875e-05, + "loss": 0.0892, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03034352362155914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016099225915968417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07678627483546734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.050852949172258376, + "weighted_rkd_loss": 0.03839313741773367 + }, + { + "epoch": 0.5862646566164154, + "flow_matching_loss": 0.05294822007417679, + "grad_norm": 0.4339646100997925, + "learning_rate": 2.9967856807370458e-05, + "loss": 0.0964, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03385459519922733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019144115131348373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08685330525040627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.05294822007417679, + "weighted_rkd_loss": 0.043426652625203135 + }, + { + "epoch": 0.6030150753768844, + "flow_matching_loss": 0.04927732534706593, + "grad_norm": 0.39898374676704407, + "learning_rate": 2.996103822930615e-05, + "loss": 0.09, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03226359337568283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01696971897035837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08149690553545952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04927732534706593, + "weighted_rkd_loss": 0.04074845276772976 + }, + { + "epoch": 0.6197654941373534, + "flow_matching_loss": 0.04757112897932529, + "grad_norm": 0.3817186653614044, + "learning_rate": 2.9953565206962653e-05, + "loss": 0.0851, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029814749024808407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015488553512841462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07511805146932601, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04757112897932529, + "weighted_rkd_loss": 0.03755902573466301 + }, + { + "epoch": 0.6365159128978225, + "flow_matching_loss": 0.04780330918729305, + "grad_norm": 0.35171931982040405, + "learning_rate": 2.9945438067234172e-05, + "loss": 0.0839, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028515780717134474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01506363358348608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07209519408643246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04780330918729305, + "weighted_rkd_loss": 0.03604759704321623 + }, + { + "epoch": 0.6532663316582915, + "flow_matching_loss": 0.04430910907685757, + "grad_norm": 0.39219847321510315, + "learning_rate": 2.9936657165628124e-05, + "loss": 0.0793, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027778214029967784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01443298477679491, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06998941153287888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04430910907685757, + "weighted_rkd_loss": 0.03499470576643944 + }, + { + "epoch": 0.6700167504187605, + "flow_matching_loss": 0.04167022332549095, + "grad_norm": 0.3763214945793152, + "learning_rate": 2.992722288624959e-05, + "loss": 0.0763, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02743688691407442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014438783749938011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06931255795061589, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04167022332549095, + "weighted_rkd_loss": 0.034656278975307944 + }, + { + "epoch": 0.6867671691792295, + "flow_matching_loss": 0.04623462222516537, + "grad_norm": 0.3273172974586487, + "learning_rate": 2.9917135641784515e-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": 0.026866826787590982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0142863011918962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06801995486021042, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04623462222516537, + "weighted_rkd_loss": 0.03400997743010521 + }, + { + "epoch": 0.7035175879396985, + "flow_matching_loss": 0.042290421202778816, + "grad_norm": 0.3280941843986511, + "learning_rate": 2.9906395873481655e-05, + "loss": 0.0747, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025736703351140023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013280625827610493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06475403234362602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042290421202778816, + "weighted_rkd_loss": 0.03237701617181301 + }, + { + "epoch": 0.7202680067001676, + "flow_matching_loss": 0.04566530957818031, + "grad_norm": 0.3375310003757477, + "learning_rate": 2.989500405113326e-05, + "loss": 0.0792, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02661526370793581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013754750974476338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06698527745902538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04566530957818031, + "weighted_rkd_loss": 0.03349263872951269 + }, + { + "epoch": 0.7370184254606366, + "flow_matching_loss": 0.041253012418746945, + "grad_norm": 0.36417585611343384, + "learning_rate": 2.9882960673054547e-05, + "loss": 0.0736, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025740651600062846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013157044537365437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06463834792375564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.041253012418746945, + "weighted_rkd_loss": 0.03231917396187782 + }, + { + "epoch": 0.7537688442211056, + "flow_matching_loss": 0.0429521806538105, + "grad_norm": 0.31173717975616455, + "learning_rate": 2.9870266266061886e-05, + "loss": 0.0734, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02408239897340536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012752699572592974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060917497426271436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0429521806538105, + "weighted_rkd_loss": 0.030458748713135718 + }, + { + "epoch": 0.7705192629815746, + "flow_matching_loss": 0.04328401610255241, + "grad_norm": 0.32832983136177063, + "learning_rate": 2.9856921385449766e-05, + "loss": 0.0763, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02588608395308256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01416186960414052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06593403704464436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04328401610255241, + "weighted_rkd_loss": 0.03296701852232218 + }, + { + "epoch": 0.7872696817420436, + "flow_matching_loss": 0.042371243983507154, + "grad_norm": 0.322933167219162, + "learning_rate": 2.9842926614966502e-05, + "loss": 0.0717, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023254065960645675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012198315933346748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0587064478546381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.042371243983507154, + "weighted_rkd_loss": 0.02935322392731905 + }, + { + "epoch": 0.8040201005025126, + "flow_matching_loss": 0.036948934011161325, + "grad_norm": 0.37090227007865906, + "learning_rate": 2.982828256678868e-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": 0.023912062868475914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012023937422782183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05984806343913078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.036948934011161325, + "weighted_rkd_loss": 0.02992403171956539 + }, + { + "epoch": 0.8207705192629816, + "flow_matching_loss": 0.04164762869477272, + "grad_norm": 0.2612155079841614, + "learning_rate": 2.9812989881494412e-05, + "loss": 0.0693, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022154004499316215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011067771259695292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05537578016519547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04164762869477272, + "weighted_rkd_loss": 0.027687890082597734 + }, + { + "epoch": 0.8375209380234506, + "flow_matching_loss": 0.041829455830156806, + "grad_norm": 0.3849196434020996, + "learning_rate": 2.9797049228035296e-05, + "loss": 0.0696, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022308938018977643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010920075606554747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05553795211017132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.041829455830156806, + "weighted_rkd_loss": 0.02776897605508566 + }, + { + "epoch": 0.8542713567839196, + "flow_matching_loss": 0.039186670817434786, + "grad_norm": 0.2965879738330841, + "learning_rate": 2.978046130370715e-05, + "loss": 0.0642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02009088359773159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009756325092166663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04993809200823307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.039186670817434786, + "weighted_rkd_loss": 0.024969046004116535 + }, + { + "epoch": 0.8710217755443886, + "flow_matching_loss": 0.03997811581939459, + "grad_norm": 0.3125465512275696, + "learning_rate": 2.9763226834119535e-05, + "loss": 0.0675, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0217224583029747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011559972260147332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05500488877296448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03997811581939459, + "weighted_rkd_loss": 0.02750244438648224 + }, + { + "epoch": 0.8877721943048577, + "flow_matching_loss": 0.037035029381513596, + "grad_norm": 0.28196239471435547, + "learning_rate": 2.9745346573163963e-05, + "loss": 0.0683, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024690970405936243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013126031681895255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06250797249376774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.037035029381513596, + "weighted_rkd_loss": 0.03125398624688387 + }, + { + "epoch": 0.9045226130653267, + "flow_matching_loss": 0.03601754531264305, + "grad_norm": 0.3579603135585785, + "learning_rate": 2.9726821302980987e-05, + "loss": 0.0653, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023435172624886036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011736051738262176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05860639661550522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03601754531264305, + "weighted_rkd_loss": 0.02930319830775261 + }, + { + "epoch": 0.9212730318257957, + "flow_matching_loss": 0.035668034479022025, + "grad_norm": 0.33571353554725647, + "learning_rate": 2.9707651833925935e-05, + "loss": 0.0603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01971091693267226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009920926159247756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049342759884893894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035668034479022025, + "weighted_rkd_loss": 0.024671379942446947 + }, + { + "epoch": 0.9380234505862647, + "flow_matching_loss": 0.04034342188388109, + "grad_norm": 0.37223154306411743, + "learning_rate": 2.968783900453348e-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": 0.02227401491254568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010327170602977276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054875200241804124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04034342188388109, + "weighted_rkd_loss": 0.027437600120902062 + }, + { + "epoch": 0.9547738693467337, + "flow_matching_loss": 0.03982900306582451, + "grad_norm": 0.3271760940551758, + "learning_rate": 2.9667383681480968e-05, + "loss": 0.0647, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019855904299765824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010042787250131368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04975459612905979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03982900306582451, + "weighted_rkd_loss": 0.024877298064529894 + }, + { + "epoch": 0.9715242881072027, + "flow_matching_loss": 0.04037307444959879, + "grad_norm": 0.3598499596118927, + "learning_rate": 2.9646286759550493e-05, + "loss": 0.0652, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01992246611043811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009732879279181362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04957781210541725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.04037307444959879, + "weighted_rkd_loss": 0.024788906052708625 + }, + { + "epoch": 0.9882747068676717, + "flow_matching_loss": 0.03921057693660259, + "grad_norm": 0.4509306848049164, + "learning_rate": 2.9624549161589764e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018727950938045978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009081155760213733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04653705805540085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03921057693660259, + "weighted_rkd_loss": 0.023268529027700425 + }, + { + "epoch": 1.0050251256281406, + "flow_matching_loss": 0.03464273381978274, + "grad_norm": 0.3406599164009094, + "learning_rate": 2.9602171838471745e-05, + "loss": 0.0614, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02158996183425188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01035784943960607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05353777259588242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03464273381978274, + "weighted_rkd_loss": 0.02676888629794121 + }, + { + "epoch": 1.0217755443886096, + "flow_matching_loss": 0.032238774560391904, + "grad_norm": 0.3566734194755554, + "learning_rate": 2.957915576905303e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019271148182451726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010775866499170662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049318162724375725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032238774560391904, + "weighted_rkd_loss": 0.024659081362187862 + }, + { + "epoch": 1.0385259631490786, + "flow_matching_loss": 0.033852733485400674, + "grad_norm": 0.3310783803462982, + "learning_rate": 2.9555501960131058e-05, + "loss": 0.0578, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019164290837943555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009481828706339002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047810410335659984, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033852733485400674, + "weighted_rkd_loss": 0.023905205167829992 + }, + { + "epoch": 1.0552763819095476, + "flow_matching_loss": 0.03569246102124453, + "grad_norm": 0.29758018255233765, + "learning_rate": 2.9531211446400068e-05, + "loss": 0.0619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020986123383045195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010503490781411528, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052475737407803535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03569246102124453, + "weighted_rkd_loss": 0.026237868703901768 + }, + { + "epoch": 1.0720268006700167, + "flow_matching_loss": 0.03429760187864304, + "grad_norm": 0.34033986926078796, + "learning_rate": 2.9506285290405813e-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": 0.019663255382329225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009310608357191085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048637119308114055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03429760187864304, + "weighted_rkd_loss": 0.024318559654057027 + }, + { + "epoch": 1.0887772194304857, + "flow_matching_loss": 0.03442421518266201, + "grad_norm": 0.3421791195869446, + "learning_rate": 2.948072458249911e-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": 0.018693058285862207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00901120575144887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04639732241630554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03442421518266201, + "weighted_rkd_loss": 0.02319866120815277 + }, + { + "epoch": 1.1055276381909547, + "flow_matching_loss": 0.033443206921219826, + "grad_norm": 0.3156006932258606, + "learning_rate": 2.9454530440788125e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020682687778025864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009997753286734223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05136312879621983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033443206921219826, + "weighted_rkd_loss": 0.025681564398109914 + }, + { + "epoch": 1.1222780569514237, + "flow_matching_loss": 0.035716149397194384, + "grad_norm": 0.3609527349472046, + "learning_rate": 2.9427704011089474e-05, + "loss": 0.0601, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01966800969094038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009427881613373757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04876390062272549, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.035716149397194384, + "weighted_rkd_loss": 0.024381950311362744 + }, + { + "epoch": 1.1390284757118927, + "flow_matching_loss": 0.03611791767179966, + "grad_norm": 0.29612571001052856, + "learning_rate": 2.94002464668781e-05, + "loss": 0.0577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017426843661814927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008403770998120308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04325745925307274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03611791767179966, + "weighted_rkd_loss": 0.02162872962653637 + }, + { + "epoch": 1.1557788944723617, + "flow_matching_loss": 0.03446404431015253, + "grad_norm": 0.26498663425445557, + "learning_rate": 2.9372159009235936e-05, + "loss": 0.0559, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01693731788545847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008955044439062476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042829679884016515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03446404431015253, + "weighted_rkd_loss": 0.021414839942008258 + }, + { + "epoch": 1.1725293132328307, + "flow_matching_loss": 0.03775186575949192, + "grad_norm": 0.2350740283727646, + "learning_rate": 2.9343442866799365e-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": 0.01552609372884035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007015585526823997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03806777279824018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03775186575949192, + "weighted_rkd_loss": 0.01903388639912009 + }, + { + "epoch": 1.1892797319932997, + "flow_matching_loss": 0.034802002087235454, + "grad_norm": 0.27234312891960144, + "learning_rate": 2.9314099295705495e-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": 0.01639019064605236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008023013034835458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04080339428037405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.034802002087235454, + "weighted_rkd_loss": 0.020401697140187025 + }, + { + "epoch": 1.2060301507537687, + "flow_matching_loss": 0.033095156960189345, + "grad_norm": 0.2403380125761032, + "learning_rate": 2.928412957953718e-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": 0.016718606650829314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00849555665627122, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04193276967853308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033095156960189345, + "weighted_rkd_loss": 0.02096638483926654 + }, + { + "epoch": 1.2227805695142377, + "flow_matching_loss": 0.03338757548481226, + "grad_norm": 0.34521374106407166, + "learning_rate": 2.925353502926689e-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": 0.016283851675689222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007864421885460614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04043212570250034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03338757548481226, + "weighted_rkd_loss": 0.02021606285125017 + }, + { + "epoch": 1.2395309882747068, + "flow_matching_loss": 0.029536503180861474, + "grad_norm": 0.313785195350647, + "learning_rate": 2.922231698319937e-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": 0.015204551909118891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007029514294117689, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03743861857801676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029536503180861474, + "weighted_rkd_loss": 0.01871930928900838 + }, + { + "epoch": 1.2562814070351758, + "flow_matching_loss": 0.031847203895449636, + "grad_norm": 0.31126895546913147, + "learning_rate": 2.9190476806913088e-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": 0.015663769003003834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007163937482982874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038491476140916345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031847203895449636, + "weighted_rkd_loss": 0.019245738070458172 + }, + { + "epoch": 1.2730318257956448, + "flow_matching_loss": 0.03111131377518177, + "grad_norm": 0.24699684977531433, + "learning_rate": 2.9158015893200504e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015276441443711519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007583758374676108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038136640936136244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03111131377518177, + "weighted_rkd_loss": 0.019068320468068122 + }, + { + "epoch": 1.2897822445561138, + "flow_matching_loss": 0.03199950084090233, + "grad_norm": 0.26863768696784973, + "learning_rate": 2.9124935662007137e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014959248062223197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007236627629026771, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03715512380003929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03199950084090233, + "weighted_rkd_loss": 0.018577561900019645 + }, + { + "epoch": 1.3065326633165828, + "flow_matching_loss": 0.031591197848320006, + "grad_norm": 0.27149420976638794, + "learning_rate": 2.9091237560369467e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01730004707351327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007396387634798884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041996481455862524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031591197848320006, + "weighted_rkd_loss": 0.020998240727931262 + }, + { + "epoch": 1.3232830820770518, + "flow_matching_loss": 0.03199263922870159, + "grad_norm": 0.23743563890457153, + "learning_rate": 2.905692306235163e-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": 0.014362255297601222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007487459061667323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03621196895837784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03199263922870159, + "weighted_rkd_loss": 0.01810598447918892 + }, + { + "epoch": 1.3400335008375208, + "flow_matching_loss": 0.032203089259564874, + "grad_norm": 0.26545020937919617, + "learning_rate": 2.9021993668980925e-05, + "loss": 0.0509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015238793008029461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006904208846390247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03738179467618465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.032203089259564874, + "weighted_rkd_loss": 0.018690897338092326 + }, + { + "epoch": 1.3567839195979898, + "flow_matching_loss": 0.03435746598988772, + "grad_norm": 0.3084009289741516, + "learning_rate": 2.8986450908182185e-05, + "loss": 0.0519, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014165364857763051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00666372743435204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03499445673078298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03435746598988772, + "weighted_rkd_loss": 0.01749722836539149 + }, + { + "epoch": 1.3735343383584588, + "flow_matching_loss": 0.03177093621343374, + "grad_norm": 0.26156696677207947, + "learning_rate": 2.8950296334710902e-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": 0.01559155946597457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0073688535019755365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03855197243392468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03177093621343374, + "weighted_rkd_loss": 0.01927598621696234 + }, + { + "epoch": 1.3902847571189278, + "flow_matching_loss": 0.03415762279182673, + "grad_norm": 0.3244900107383728, + "learning_rate": 2.8913531530085248e-05, + "loss": 0.054, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015872035082429646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00787778990343213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03962186016142368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03415762279182673, + "weighted_rkd_loss": 0.01981093008071184 + }, + { + "epoch": 1.4070351758793969, + "flow_matching_loss": 0.031543026119470595, + "grad_norm": 0.24670489132404327, + "learning_rate": 2.887615810251687e-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": 0.014911935105919838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007098326459527016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03692219667136669, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031543026119470595, + "weighted_rkd_loss": 0.018461098335683344 + }, + { + "epoch": 1.4237855946398659, + "flow_matching_loss": 0.030299522541463376, + "grad_norm": 0.29750338196754456, + "learning_rate": 2.8838177686840577e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013420206494629383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006395753938704729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033236166834831236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030299522541463376, + "weighted_rkd_loss": 0.016618083417415618 + }, + { + "epoch": 1.4405360134003349, + "flow_matching_loss": 0.03428787123411894, + "grad_norm": 0.2844262421131134, + "learning_rate": 2.8799591944442776e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014772494789212942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007544750696979463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0370897401124239, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03428787123411894, + "weighted_rkd_loss": 0.01854487005621195 + }, + { + "epoch": 1.457286432160804, + "flow_matching_loss": 0.028464125469326973, + "grad_norm": 0.2586044669151306, + "learning_rate": 2.8760402563188848e-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": 0.015234122145920992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007229142403230071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03769738636910915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028464125469326973, + "weighted_rkd_loss": 0.018848693184554576 + }, + { + "epoch": 1.474036850921273, + "flow_matching_loss": 0.03240107372403145, + "grad_norm": 0.3451785743236542, + "learning_rate": 2.8720611257349285e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01441895104944706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007548008835874498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036385910399258134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03240107372403145, + "weighted_rkd_loss": 0.018192955199629067 + }, + { + "epoch": 1.490787269681742, + "flow_matching_loss": 0.031261354126036166, + "grad_norm": 0.2677576243877411, + "learning_rate": 2.868021976752471e-05, + "loss": 0.0504, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015447228122502566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007448036875575781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03834249284118414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.031261354126036166, + "weighted_rkd_loss": 0.01917124642059207 + }, + { + "epoch": 1.507537688442211, + "flow_matching_loss": 0.029524659179151057, + "grad_norm": 0.2589884400367737, + "learning_rate": 2.8639229860569738e-05, + "loss": 0.0467, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013584412820637225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007130756136029958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03429958187043667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029524659179151057, + "weighted_rkd_loss": 0.017149790935218335 + }, + { + "epoch": 1.52428810720268, + "flow_matching_loss": 0.030357540026307105, + "grad_norm": 0.2322537899017334, + "learning_rate": 2.8597643329515688e-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": 0.015393463242799044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007288130559027195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03807505648583174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030357540026307105, + "weighted_rkd_loss": 0.01903752824291587 + }, + { + "epoch": 1.541038525963149, + "flow_matching_loss": 0.033645805157721045, + "grad_norm": 0.2303711622953415, + "learning_rate": 2.855546199349214e-05, + "loss": 0.0525, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0153488309122622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070605908520519735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037758252769708636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.033645805157721045, + "weighted_rkd_loss": 0.018879126384854318 + }, + { + "epoch": 1.557788944723618, + "flow_matching_loss": 0.029417644068598748, + "grad_norm": 0.33168652653694153, + "learning_rate": 2.85126876976474e-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": 0.013482074066996575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007291349256411195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03425549734383822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.029417644068598748, + "weighted_rkd_loss": 0.01712774867191911 + }, + { + "epoch": 1.574539363484087, + "flow_matching_loss": 0.02936752066016197, + "grad_norm": 0.32166430354118347, + "learning_rate": 2.8469322313067726e-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": 0.014363228995352983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006481660483404994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035208118706941606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02936752066016197, + "weighted_rkd_loss": 0.017604059353470803 + }, + { + "epoch": 1.591289782244556, + "flow_matching_loss": 0.030744343996047974, + "grad_norm": 0.2652220129966736, + "learning_rate": 2.8425367736695522e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012823993526399135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006567410938441753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03221539761871099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.030744343996047974, + "weighted_rkd_loss": 0.016107698809355496 + }, + { + "epoch": 1.608040201005025, + "flow_matching_loss": 0.027458759397268294, + "grad_norm": 0.2823118269443512, + "learning_rate": 2.838082589124636e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014246038906276227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006480111461132765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03497218918055296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027458759397268294, + "weighted_rkd_loss": 0.01748609459027648 + }, + { + "epoch": 1.624790619765494, + "flow_matching_loss": 0.02682869080454111, + "grad_norm": 0.27871882915496826, + "learning_rate": 2.8335698725124843e-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": 0.015229499991983175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0072166129481047395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037675613351166246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02682869080454111, + "weighted_rkd_loss": 0.018837806675583123 + }, + { + "epoch": 1.641541038525963, + "flow_matching_loss": 0.027610163576900958, + "grad_norm": 0.27080997824668884, + "learning_rate": 2.8289988212339422e-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": 0.014654815848916768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007433690736070276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03674332220107317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027610163576900958, + "weighted_rkd_loss": 0.018371661100536583 + }, + { + "epoch": 1.658291457286432, + "flow_matching_loss": 0.026477544382214547, + "grad_norm": 0.29053348302841187, + "learning_rate": 2.8243696352415997e-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": 0.014315103739500045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006944430945441127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035574638098478314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026477544382214547, + "weighted_rkd_loss": 0.017787319049239157 + }, + { + "epoch": 1.675041876046901, + "flow_matching_loss": 0.03149757422506809, + "grad_norm": 0.24316534399986267, + "learning_rate": 2.8196825170310495e-05, + "loss": 0.0489, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013985770475119352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006879748497158289, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034851289354264736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.03149757422506809, + "weighted_rkd_loss": 0.017425644677132368 + }, + { + "epoch": 1.69179229480737, + "flow_matching_loss": 0.028719986975193023, + "grad_norm": 0.2793906033039093, + "learning_rate": 2.8149376716320256e-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": 0.014713972900062799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007857491169124842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03728543724864721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028719986975193023, + "weighted_rkd_loss": 0.018642718624323606 + }, + { + "epoch": 1.708542713567839, + "flow_matching_loss": 0.02652828861027956, + "grad_norm": 0.24760262668132782, + "learning_rate": 2.8101353065994373e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015948845073580742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008608205150812865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04050589501857758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02652828861027956, + "weighted_rkd_loss": 0.02025294750928879 + }, + { + "epoch": 1.725293132328308, + "flow_matching_loss": 0.02748129330575466, + "grad_norm": 0.2966518998146057, + "learning_rate": 2.8052756320042885e-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": 0.013009252492338419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00621428107842803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03223278634250164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02748129330575466, + "weighted_rkd_loss": 0.01611639317125082 + }, + { + "epoch": 1.742043551088777, + "flow_matching_loss": 0.02748612593859434, + "grad_norm": 0.2400917261838913, + "learning_rate": 2.8003588604244893e-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": 0.012675477657467127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006017118878662586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031368074007332325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02748612593859434, + "weighted_rkd_loss": 0.015684037003666162 + }, + { + "epoch": 1.758793969849246, + "flow_matching_loss": 0.0283775020390749, + "grad_norm": 0.23072923719882965, + "learning_rate": 2.7953852069355565e-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": 0.012886785715818406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007881310116499663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03365488145500421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0283775020390749, + "weighted_rkd_loss": 0.016827440727502106 + }, + { + "epoch": 1.775544388609715, + "flow_matching_loss": 0.028830614313483237, + "grad_norm": 0.19938981533050537, + "learning_rate": 2.790354889101206e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011350309383124113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005617117416113615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028317735902965067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028830614313483237, + "weighted_rkd_loss": 0.014158867951482534 + }, + { + "epoch": 1.792294807370184, + "flow_matching_loss": 0.02743749525398016, + "grad_norm": 0.20770491659641266, + "learning_rate": 2.7852681269638366e-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": 0.013101563602685929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006368691730313003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03257181905210018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02743749525398016, + "weighted_rkd_loss": 0.01628590952605009 + }, + { + "epoch": 1.809045226130653, + "flow_matching_loss": 0.025509939529001713, + "grad_norm": 0.23055991530418396, + "learning_rate": 2.780125143034902e-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": 0.01217031548731029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005663800402544439, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030004431121051313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025509939529001713, + "weighted_rkd_loss": 0.015002215560525656 + }, + { + "epoch": 1.8257956448911221, + "flow_matching_loss": 0.027603413537144662, + "grad_norm": 0.24784690141677856, + "learning_rate": 2.7749261622851812e-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": 0.012131767626851798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005731426505371928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02999496180564165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027603413537144662, + "weighted_rkd_loss": 0.014997480902820826 + }, + { + "epoch": 1.8425460636515911, + "flow_matching_loss": 0.02833728678524494, + "grad_norm": 0.2339480221271515, + "learning_rate": 2.7696714121349335e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011987958382815123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005508459242992103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029484376311302185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02833728678524494, + "weighted_rkd_loss": 0.014742188155651093 + }, + { + "epoch": 1.8592964824120601, + "flow_matching_loss": 0.028176490403711795, + "grad_norm": 0.21943725645542145, + "learning_rate": 2.764361122443954e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012409442663192749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006674071168527007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03149295635521412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028176490403711795, + "weighted_rkd_loss": 0.01574647817760706 + }, + { + "epoch": 1.8760469011725294, + "flow_matching_loss": 0.02386368773877621, + "grad_norm": 0.28784745931625366, + "learning_rate": 2.758995525501517e-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": 0.012525459285825491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006127721210941672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031178639829158784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02386368773877621, + "weighted_rkd_loss": 0.015589319914579392 + }, + { + "epoch": 1.8927973199329984, + "flow_matching_loss": 0.026401301100850105, + "grad_norm": 0.302274614572525, + "learning_rate": 2.753574856016215e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012502617854624986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005733438022434712, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03073867317289114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026401301100850105, + "weighted_rkd_loss": 0.01536933658644557 + }, + { + "epoch": 1.9095477386934674, + "flow_matching_loss": 0.025203173607587816, + "grad_norm": 0.22959397733211517, + "learning_rate": 2.7480993511056913e-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": 0.012079661898314952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005447410838678479, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029606734588742255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025203173607587816, + "weighted_rkd_loss": 0.014803367294371128 + }, + { + "epoch": 1.9262981574539364, + "flow_matching_loss": 0.02645403500646353, + "grad_norm": 0.20805855095386505, + "learning_rate": 2.7425692502862684e-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": 0.012591348122805357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005284116393886507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030466812662780285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02645403500646353, + "weighted_rkd_loss": 0.015233406331390142 + }, + { + "epoch": 1.9430485762144054, + "flow_matching_loss": 0.028687219321727752, + "grad_norm": 0.23209843039512634, + "learning_rate": 2.7369847954624728e-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": 0.01346703628078103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005913856625556946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032847929000854495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028687219321727752, + "weighted_rkd_loss": 0.016423964500427247 + }, + { + "epoch": 1.9597989949748744, + "flow_matching_loss": 0.0267184529453516, + "grad_norm": 0.20264476537704468, + "learning_rate": 2.731346230916448e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012482394743710757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005907515040598809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030872304178774356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0267184529453516, + "weighted_rkd_loss": 0.015436152089387178 + }, + { + "epoch": 1.9765494137353434, + "flow_matching_loss": 0.02526174336671829, + "grad_norm": 0.2656524181365967, + "learning_rate": 2.7256538032972746e-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": 0.011073235142976045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004903874779120088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02705034501850605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02526174336671829, + "weighted_rkd_loss": 0.013525172509253025 + }, + { + "epoch": 1.9932998324958124, + "flow_matching_loss": 0.024283120594918727, + "grad_norm": 0.18750469386577606, + "learning_rate": 2.7199077616101773e-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": 0.010565282870084047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005318943597376347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026449509151279926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024283120594918727, + "weighted_rkd_loss": 0.013224754575639963 + }, + { + "epoch": 2.0100502512562812, + "flow_matching_loss": 0.023823299817740916, + "grad_norm": 0.28387999534606934, + "learning_rate": 2.714108357205634e-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": 0.01328891459852457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061988724861294035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032776701636612414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023823299817740916, + "weighted_rkd_loss": 0.016388350818306207 + }, + { + "epoch": 2.0268006700167502, + "flow_matching_loss": 0.027536653354763983, + "grad_norm": 0.23609916865825653, + "learning_rate": 2.7082558437683803e-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": 0.010250479681417346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046427268767729405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025143686309456827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027536653354763983, + "weighted_rkd_loss": 0.012571843154728413 + }, + { + "epoch": 2.0435510887772192, + "flow_matching_loss": 0.025359765440225602, + "grad_norm": 0.2570204436779022, + "learning_rate": 2.702350477306315e-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": 0.012231682194396853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005420093261636794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02988345790654421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025359765440225602, + "weighted_rkd_loss": 0.014941728953272105 + }, + { + "epoch": 2.0603015075376883, + "flow_matching_loss": 0.02634656634181738, + "grad_norm": 0.22433871030807495, + "learning_rate": 2.6963925161392963e-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": 0.010704432427883149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047349676489830015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026143832691013814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02634656634181738, + "weighted_rkd_loss": 0.013071916345506907 + }, + { + "epoch": 2.0770519262981573, + "flow_matching_loss": 0.02522501554340124, + "grad_norm": 0.2080315798521042, + "learning_rate": 2.6903822208878476e-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": 0.010163325816392899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004687550477683544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02501420211046934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02522501554340124, + "weighted_rkd_loss": 0.01250710105523467 + }, + { + "epoch": 2.0938023450586263, + "flow_matching_loss": 0.02463902086019516, + "grad_norm": 0.25818106532096863, + "learning_rate": 2.6843198544617535e-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": 0.012039071694016457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006278580939397216, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030356723815202713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02463902086019516, + "weighted_rkd_loss": 0.015178361907601357 + }, + { + "epoch": 2.1105527638190953, + "flow_matching_loss": 0.027839330770075322, + "grad_norm": 0.2640082538127899, + "learning_rate": 2.6782056820485612e-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": 0.012098926212638616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005462716426700353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029660568572580814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.027839330770075322, + "weighted_rkd_loss": 0.014830284286290407 + }, + { + "epoch": 2.1273031825795643, + "flow_matching_loss": 0.028276474215090275, + "grad_norm": 0.295583575963974, + "learning_rate": 2.672039971101979e-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": 0.013212186470627784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006356807774864137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032781181018799545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.028276474215090275, + "weighted_rkd_loss": 0.016390590509399772 + }, + { + "epoch": 2.1440536013400333, + "flow_matching_loss": 0.026197852566838264, + "grad_norm": 0.20729024708271027, + "learning_rate": 2.665822991330178e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01121538388542831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005526613956317305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027957382053136824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.026197852566838264, + "weighted_rkd_loss": 0.013978691026568412 + }, + { + "epoch": 2.1608040201005023, + "flow_matching_loss": 0.02445336002856493, + "grad_norm": 0.19470828771591187, + "learning_rate": 2.6595550146839934e-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": 0.011698291730135679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005225977627560496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02862256094813347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02445336002856493, + "weighted_rkd_loss": 0.014311280474066735 + }, + { + "epoch": 2.1775544388609713, + "flow_matching_loss": 0.02761701140552759, + "grad_norm": 0.29102823138237, + "learning_rate": 2.6532363153450287e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012762863840907811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00593646764755249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031462194956839086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02761701140552759, + "weighted_rkd_loss": 0.015731097478419543 + }, + { + "epoch": 2.1943048576214403, + "flow_matching_loss": 0.023142579942941666, + "grad_norm": 0.2832508087158203, + "learning_rate": 2.6468671697136634e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01395815759897232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007116776378825307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03503309153020382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023142579942941666, + "weighted_rkd_loss": 0.01751654576510191 + }, + { + "epoch": 2.2110552763819094, + "flow_matching_loss": 0.024112015962600708, + "grad_norm": 0.22076784074306488, + "learning_rate": 2.640447856396961e-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": 0.010843491833657026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005128723871894181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026815707236528395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024112015962600708, + "weighted_rkd_loss": 0.013407853618264198 + }, + { + "epoch": 2.2278056951423784, + "flow_matching_loss": 0.02480712588876486, + "grad_norm": 0.24121958017349243, + "learning_rate": 2.6339786561964805e-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": 0.010331244813278318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004812800046056509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025475289672613144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02480712588876486, + "weighted_rkd_loss": 0.012737644836306572 + }, + { + "epoch": 2.2445561139028474, + "flow_matching_loss": 0.02151898145675659, + "grad_norm": 0.2851496636867523, + "learning_rate": 2.6274598520959966e-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": 0.011607506917789579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005171733908355236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028386748023331167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02151898145675659, + "weighted_rkd_loss": 0.014193374011665583 + }, + { + "epoch": 2.2613065326633164, + "flow_matching_loss": 0.02450380623340607, + "grad_norm": 0.19669336080551147, + "learning_rate": 2.6208917292491185e-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": 0.011534726107493043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005064515001140535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02813396751880646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02450380623340607, + "weighted_rkd_loss": 0.01406698375940323 + }, + { + "epoch": 2.2780569514237854, + "flow_matching_loss": 0.0243028256110847, + "grad_norm": 0.22813156247138977, + "learning_rate": 2.6142745749668176e-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": 0.008905993402004242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004185553826391697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021997540444135665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0243028256110847, + "weighted_rkd_loss": 0.010998770222067833 + }, + { + "epoch": 2.2948073701842544, + "flow_matching_loss": 0.023458981513977052, + "grad_norm": 0.1727350801229477, + "learning_rate": 2.607608678704859e-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": 0.010793544258922339, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005353818205185235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02694090660661459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023458981513977052, + "weighted_rkd_loss": 0.013470453303307294 + }, + { + "epoch": 2.3115577889447234, + "flow_matching_loss": 0.02446927558630705, + "grad_norm": 0.20639468729496002, + "learning_rate": 2.600894332051139e-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": 0.01072792410850525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047226678114384415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026178515702486038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02446927558630705, + "weighted_rkd_loss": 0.013089257851243019 + }, + { + "epoch": 2.3283082077051924, + "flow_matching_loss": 0.02812897562980652, + "grad_norm": 0.23591165244579315, + "learning_rate": 2.594131828712932e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010303571028634906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005087628937326371, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02569477092474699, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02812897562980652, + "weighted_rkd_loss": 0.012847385462373495 + }, + { + "epoch": 2.3450586264656614, + "flow_matching_loss": 0.023055185005068778, + "grad_norm": 0.2851240336894989, + "learning_rate": 2.5873214645040414e-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": 0.011493522580713034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053280909080058335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028315136022865773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023055185005068778, + "weighted_rkd_loss": 0.014157568011432886 + }, + { + "epoch": 2.3618090452261304, + "flow_matching_loss": 0.023752907849848272, + "grad_norm": 0.2843516170978546, + "learning_rate": 2.5804635373318604e-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": 0.012701206561177968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00591087422799319, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03131328746676445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023752907849848272, + "weighted_rkd_loss": 0.015656643733382224 + }, + { + "epoch": 2.3785594639865995, + "flow_matching_loss": 0.024771238677203655, + "grad_norm": 0.23561623692512512, + "learning_rate": 2.57355834718434e-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": 0.008858569525182247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004208746063522994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02192588523030281, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024771238677203655, + "weighted_rkd_loss": 0.010962942615151405 + }, + { + "epoch": 2.3953098827470685, + "flow_matching_loss": 0.024008012004196645, + "grad_norm": 0.20277637243270874, + "learning_rate": 2.566606196116867e-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": 0.01078841257840395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050114454235881565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0265882708132267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024008012004196645, + "weighted_rkd_loss": 0.01329413540661335 + }, + { + "epoch": 2.4120603015075375, + "flow_matching_loss": 0.02423899285495281, + "grad_norm": 0.21418488025665283, + "learning_rate": 2.559607388239051e-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": 0.010564777301624417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005194381531327963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02632393632084131, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02423899285495281, + "weighted_rkd_loss": 0.013161968160420656 + }, + { + "epoch": 2.4288107202680065, + "flow_matching_loss": 0.021590460650622846, + "grad_norm": 0.23149818181991577, + "learning_rate": 2.552562229701422e-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": 0.010688819037750364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005056107765994966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026433745771646498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021590460650622846, + "weighted_rkd_loss": 0.013216872885823249 + }, + { + "epoch": 2.4455611390284755, + "flow_matching_loss": 0.0240889685228467, + "grad_norm": 0.1726461946964264, + "learning_rate": 2.545471028682036e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009562594909220934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043014321010559796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023426621593534945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0240889685228467, + "weighted_rkd_loss": 0.011713310796767473 + }, + { + "epoch": 2.4623115577889445, + "flow_matching_loss": 0.025627955421805383, + "grad_norm": 0.2549595236778259, + "learning_rate": 2.538334095373e-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": 0.010514754336327314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004727694508619607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025757203251123427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025627955421805383, + "weighted_rkd_loss": 0.012878601625561713 + }, + { + "epoch": 2.4790619765494135, + "flow_matching_loss": 0.024999511800706385, + "grad_norm": 0.20995697379112244, + "learning_rate": 2.531151741966896e-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": 0.00883528389967978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004498904151841998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02216947190463543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024999511800706385, + "weighted_rkd_loss": 0.011084735952317714 + }, + { + "epoch": 2.4958123953098825, + "flow_matching_loss": 0.024381516501307487, + "grad_norm": 0.22541256248950958, + "learning_rate": 2.5239242826431302e-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": 0.009589857840910553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004377200687304139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023556916601955892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024381516501307487, + "weighted_rkd_loss": 0.011778458300977946 + }, + { + "epoch": 2.5125628140703515, + "flow_matching_loss": 0.02520757894963026, + "grad_norm": 0.2073884904384613, + "learning_rate": 2.5166520335541865e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010272404458373786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051223641028627755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02566717304289341, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02520757894963026, + "weighted_rkd_loss": 0.012833586521446705 + }, + { + "epoch": 2.5293132328308205, + "flow_matching_loss": 0.02480332348495722, + "grad_norm": 0.22228868305683136, + "learning_rate": 2.5093353128117996e-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": 0.008547855308279395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004206642787903547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021302353404462338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02480332348495722, + "weighted_rkd_loss": 0.010651176702231169 + }, + { + "epoch": 2.5460636515912896, + "flow_matching_loss": 0.023200121708214283, + "grad_norm": 0.1978033035993576, + "learning_rate": 2.501974440473037e-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": 0.010471200989559292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004926082771271467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025868484750390053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023200121708214283, + "weighted_rkd_loss": 0.012934242375195026 + }, + { + "epoch": 2.5628140703517586, + "flow_matching_loss": 0.02555184094235301, + "grad_norm": 0.22602304816246033, + "learning_rate": 2.4945697385263015e-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": 0.008761562407016754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039977532578632236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021520878095179795, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02555184094235301, + "weighted_rkd_loss": 0.010760439047589897 + }, + { + "epoch": 2.5795644891122276, + "flow_matching_loss": 0.021686122007668018, + "grad_norm": 0.1781231164932251, + "learning_rate": 2.487121530877243e-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": 0.010080217849463224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00466340493876487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024823840521275998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021686122007668018, + "weighted_rkd_loss": 0.012411920260637999 + }, + { + "epoch": 2.5963149078726966, + "flow_matching_loss": 0.022586009837687014, + "grad_norm": 0.21403099596500397, + "learning_rate": 2.4796301433345943e-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": 0.010508734360337257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005213955650106072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02623142395168543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022586009837687014, + "weighted_rkd_loss": 0.013115711975842714 + }, + { + "epoch": 2.6130653266331656, + "flow_matching_loss": 0.023120622523128985, + "grad_norm": 0.18488554656505585, + "learning_rate": 2.4720959035959148e-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": 0.008027114858850836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003637227532453835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019691457133740188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023120622523128985, + "weighted_rkd_loss": 0.009845728566870094 + }, + { + "epoch": 2.6298157453936346, + "flow_matching_loss": 0.02213309984654188, + "grad_norm": 0.2590029537677765, + "learning_rate": 2.464519141233257e-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": 0.008783003455027938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004119690437801182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021685697697103025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02213309984654188, + "weighted_rkd_loss": 0.010842848848551512 + }, + { + "epoch": 2.6465661641541036, + "flow_matching_loss": 0.023288248293101787, + "grad_norm": 0.20496879518032074, + "learning_rate": 2.456900187678754e-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": 0.009698561951518058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004631320876069367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024028444662690163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023288248293101787, + "weighted_rkd_loss": 0.012014222331345081 + }, + { + "epoch": 2.6633165829145726, + "flow_matching_loss": 0.02162441601976752, + "grad_norm": 0.1816551685333252, + "learning_rate": 2.4492393762101147e-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": 0.009497049171477557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004680510167963803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023674608673900366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02162441601976752, + "weighted_rkd_loss": 0.011837304336950183 + }, + { + "epoch": 2.6800670016750416, + "flow_matching_loss": 0.02216140553355217, + "grad_norm": 0.1991337537765503, + "learning_rate": 2.4415370419360508e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009966886276379228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004221933707594871, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024155706353485584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02216140553355217, + "weighted_rkd_loss": 0.012077853176742792 + }, + { + "epoch": 2.6968174204355106, + "flow_matching_loss": 0.02192650279030204, + "grad_norm": 0.2767055034637451, + "learning_rate": 2.4337935217816157e-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": 0.010798801574856043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047045488143339755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02630215175449848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02192650279030204, + "weighted_rkd_loss": 0.01315107587724924 + }, + { + "epoch": 2.7135678391959797, + "flow_matching_loss": 0.02603024896234274, + "grad_norm": 0.20454247295856476, + "learning_rate": 2.4260091544734665e-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": 0.010186881572008134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004692598734982312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025066361762583256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02603024896234274, + "weighted_rkd_loss": 0.012533180881291628 + }, + { + "epoch": 2.7303182579564487, + "flow_matching_loss": 0.025671250931918622, + "grad_norm": 0.20730584859848022, + "learning_rate": 2.4181842805250466e-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": 0.008627374796196819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042751003755256535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02152985017746687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025671250931918622, + "weighted_rkd_loss": 0.010764925088733435 + }, + { + "epoch": 2.7470686767169177, + "flow_matching_loss": 0.02359693106263876, + "grad_norm": 0.188655287027359, + "learning_rate": 2.4103192422216923e-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": 0.010495147341862321, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005308826128020883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02629912067204714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02359693106263876, + "weighted_rkd_loss": 0.01314956033602357 + }, + { + "epoch": 2.7638190954773867, + "flow_matching_loss": 0.023337428458034992, + "grad_norm": 0.23929449915885925, + "learning_rate": 2.402414383605658e-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": 0.010052195470780134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004961214587092399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025065605714917182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.023337428458034992, + "weighted_rkd_loss": 0.012532802857458591 + }, + { + "epoch": 2.7805695142378557, + "flow_matching_loss": 0.02219377104192972, + "grad_norm": 0.2548065185546875, + "learning_rate": 2.3944700504610674e-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": 0.010901725804433226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045725111151114104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026375962886959314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02219377104192972, + "weighted_rkd_loss": 0.013187981443479657 + }, + { + "epoch": 2.7973199329983247, + "flow_matching_loss": 0.021554847713559866, + "grad_norm": 0.19051788747310638, + "learning_rate": 2.3864865902987883e-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": 0.009427066240459681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004298139782622457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023152272403240203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021554847713559866, + "weighted_rkd_loss": 0.011576136201620102 + }, + { + "epoch": 2.8140703517587937, + "flow_matching_loss": 0.025536979362368584, + "grad_norm": 0.21837979555130005, + "learning_rate": 2.378464352341231e-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": 0.010489001916721464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051647582557052376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026142761670053007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.025536979362368584, + "weighted_rkd_loss": 0.013071380835026503 + }, + { + "epoch": 2.8308207705192627, + "flow_matching_loss": 0.020257471688091756, + "grad_norm": 0.21359702944755554, + "learning_rate": 2.3704036875070728e-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": 0.008906964818015695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004341627052053809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022155556827783585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020257471688091756, + "weighted_rkd_loss": 0.011077778413891792 + }, + { + "epoch": 2.8475711892797317, + "flow_matching_loss": 0.022737276274710894, + "grad_norm": 0.18399080634117126, + "learning_rate": 2.362304948395906e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009278100775554776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00438445636536926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022940658032894135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022737276274710894, + "weighted_rkd_loss": 0.011470329016447067 + }, + { + "epoch": 2.8643216080402008, + "flow_matching_loss": 0.02296969071030617, + "grad_norm": 0.2073369175195694, + "learning_rate": 2.3541684892728157e-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": 0.008697510184720158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004495504498481751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02189052477478981, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02296969071030617, + "weighted_rkd_loss": 0.010945262387394905 + }, + { + "epoch": 2.8810720268006698, + "flow_matching_loss": 0.02656156402081251, + "grad_norm": 0.2169615775346756, + "learning_rate": 2.345994666052883e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009356808802112936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004665453173220158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02337907087057829, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02656156402081251, + "weighted_rkd_loss": 0.011689535435289144 + }, + { + "epoch": 2.8978224455611388, + "flow_matching_loss": 0.02025458849966526, + "grad_norm": 0.19369857013225555, + "learning_rate": 2.337783836285614e-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": 0.009551549004390836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004140716185793281, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023243813868612052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02025458849966526, + "weighted_rkd_loss": 0.011621906934306026 + }, + { + "epoch": 2.914572864321608, + "flow_matching_loss": 0.020585576817393304, + "grad_norm": 0.2022498995065689, + "learning_rate": 2.3295363591393026e-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": 0.008104678569361567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003487278800457716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019696636125445366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020585576817393304, + "weighted_rkd_loss": 0.009848318062722683 + }, + { + "epoch": 2.931323283082077, + "flow_matching_loss": 0.02213856503367424, + "grad_norm": 0.22983857989311218, + "learning_rate": 2.3212525953853174e-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": 0.009665360115468502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004478284833021462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023809005320072175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02213856503367424, + "weighted_rkd_loss": 0.011904502660036087 + }, + { + "epoch": 2.948073701842546, + "flow_matching_loss": 0.02260482348501682, + "grad_norm": 0.2563987076282501, + "learning_rate": 2.3129329073823206e-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": 0.008846775069832802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004327321448363364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02202087165787816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02260482348501682, + "weighted_rkd_loss": 0.01101043582893908 + }, + { + "epoch": 2.964824120603015, + "flow_matching_loss": 0.024081751145422457, + "grad_norm": 0.27132245898246765, + "learning_rate": 2.3045776590604175e-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": 0.011177179915830494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004908815934322774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027263175696134567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024081751145422457, + "weighted_rkd_loss": 0.013631587848067284 + }, + { + "epoch": 2.981574539363484, + "flow_matching_loss": 0.02145145069807768, + "grad_norm": 0.23230737447738647, + "learning_rate": 2.296187215905237e-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": 0.00857270359992981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039184731664136056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021063880529254675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02145145069807768, + "weighted_rkd_loss": 0.010531940264627337 + }, + { + "epoch": 2.998324958123953, + "flow_matching_loss": 0.022610493656247854, + "grad_norm": 0.21160763502120972, + "learning_rate": 2.2877619449419437e-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": 0.00877288361079991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040158325573429465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02156159970909357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022610493656247854, + "weighted_rkd_loss": 0.010780799854546785 + }, + { + "epoch": 3.0150753768844223, + "flow_matching_loss": 0.01961312396451831, + "grad_norm": 0.21592052280902863, + "learning_rate": 2.2793022147191843e-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": 0.008552303677424789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004099966096691787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021204573288559913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01961312396451831, + "weighted_rkd_loss": 0.010602286644279956 + }, + { + "epoch": 3.0318257956448913, + "flow_matching_loss": 0.021133813075721264, + "grad_norm": 0.2408725470304489, + "learning_rate": 2.2708083952929634e-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": 0.009924392122775317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004459633771330118, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024308417923748493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021133813075721264, + "weighted_rkd_loss": 0.012154208961874247 + }, + { + "epoch": 3.0485762144053603, + "flow_matching_loss": 0.021858297660946845, + "grad_norm": 0.21889373660087585, + "learning_rate": 2.2622808582104594e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007935315230861306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036201345268636943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019490764942020178, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021858297660946845, + "weighted_rkd_loss": 0.009745382471010089 + }, + { + "epoch": 3.0653266331658293, + "flow_matching_loss": 0.02066167173907161, + "grad_norm": 0.2489049881696701, + "learning_rate": 2.2537199764937703e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008639211347326637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042247757315635685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021503198333084582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02066167173907161, + "weighted_rkd_loss": 0.010751599166542291 + }, + { + "epoch": 3.0820770519262983, + "flow_matching_loss": 0.021619039587676525, + "grad_norm": 0.18713706731796265, + "learning_rate": 2.245126124623595e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00882501546293497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004201194737106562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021851225569844247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021619039587676525, + "weighted_rkd_loss": 0.010925612784922123 + }, + { + "epoch": 3.0988274706867673, + "flow_matching_loss": 0.0222555180080235, + "grad_norm": 0.16398872435092926, + "learning_rate": 2.2364996785228547e-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": 0.008472857438027859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038555797887966036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020801294595003128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0222555180080235, + "weighted_rkd_loss": 0.010400647297501564 + }, + { + "epoch": 3.1155778894472363, + "flow_matching_loss": 0.02233735201880336, + "grad_norm": 0.195217564702034, + "learning_rate": 2.227841015540248e-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": 0.00930324881337583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003956924844533205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022563422471284865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02233735201880336, + "weighted_rkd_loss": 0.011281711235642432 + }, + { + "epoch": 3.1323283082077054, + "flow_matching_loss": 0.021456127893179654, + "grad_norm": 0.24798332154750824, + "learning_rate": 2.2191505144337433e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00950444871559739, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004383202618919313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023392099887132645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021456127893179654, + "weighted_rkd_loss": 0.011696049943566322 + }, + { + "epoch": 3.1490787269681744, + "flow_matching_loss": 0.020449751988053322, + "grad_norm": 0.21128728985786438, + "learning_rate": 2.210428555354013e-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": 0.007659428799524903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003622169536538422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018941027391701937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020449751988053322, + "weighted_rkd_loss": 0.009470513695850969 + }, + { + "epoch": 3.1658291457286434, + "flow_matching_loss": 0.020440852362662555, + "grad_norm": 0.19786065816879272, + "learning_rate": 2.201675519827802e-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": 0.00879075680859387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00397156581748277, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02155307959765196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020440852362662555, + "weighted_rkd_loss": 0.01077653979882598 + }, + { + "epoch": 3.1825795644891124, + "flow_matching_loss": 0.021140540950000286, + "grad_norm": 0.17299114167690277, + "learning_rate": 2.1928917907412403e-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": 0.008886780729517341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004150486085563898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021924047172069548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021140540950000286, + "weighted_rkd_loss": 0.010962023586034774 + }, + { + "epoch": 3.1993299832495814, + "flow_matching_loss": 0.021827292442321778, + "grad_norm": 0.29607075452804565, + "learning_rate": 2.1840777523230937e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008809508057311177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038985042599961163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021517520304769278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021827292442321778, + "weighted_rkd_loss": 0.010758760152384639 + }, + { + "epoch": 3.2160804020100504, + "flow_matching_loss": 0.02189473919570446, + "grad_norm": 0.21935483813285828, + "learning_rate": 2.1752337901279552e-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": 0.008246973948553204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004177061165682972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02067100889980793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02189473919570446, + "weighted_rkd_loss": 0.010335504449903965 + }, + { + "epoch": 3.2328308207705194, + "flow_matching_loss": 0.02115060854703188, + "grad_norm": 0.21308168768882751, + "learning_rate": 2.166360291019383e-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": 0.008105326350778342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034979252610355615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01970857810229063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02115060854703188, + "weighted_rkd_loss": 0.009854289051145316 + }, + { + "epoch": 3.2495812395309884, + "flow_matching_loss": 0.019409695640206337, + "grad_norm": 0.2219625562429428, + "learning_rate": 2.157457643152973e-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": 0.007863681856542825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037062383722513913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01943360213190317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019409695640206337, + "weighted_rkd_loss": 0.009716801065951585 + }, + { + "epoch": 3.2663316582914574, + "flow_matching_loss": 0.02243496719747782, + "grad_norm": 0.17238378524780273, + "learning_rate": 2.148526235959385e-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": 0.008457799488678575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003480206034146249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020395804755389692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02243496719747782, + "weighted_rkd_loss": 0.010197902377694846 + }, + { + "epoch": 3.2830820770519265, + "flow_matching_loss": 0.021297021582722665, + "grad_norm": 0.21285296976566315, + "learning_rate": 2.1395664601273016e-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": 0.008106729062274099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003673051320947707, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019886509422212838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021297021582722665, + "weighted_rkd_loss": 0.009943254711106419 + }, + { + "epoch": 3.2998324958123955, + "flow_matching_loss": 0.02030411781743169, + "grad_norm": 0.22372476756572723, + "learning_rate": 2.130578707586345e-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": 0.008725496334955096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004435207205824554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021886199712753296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02030411781743169, + "weighted_rkd_loss": 0.010943099856376648 + }, + { + "epoch": 3.3165829145728645, + "flow_matching_loss": 0.021738264430314302, + "grad_norm": 0.20458626747131348, + "learning_rate": 2.1215633714899264e-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": 0.008855098765343428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004159924946725369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021870122756808996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021738264430314302, + "weighted_rkd_loss": 0.010935061378404498 + }, + { + "epoch": 3.3333333333333335, + "flow_matching_loss": 0.022326320968568326, + "grad_norm": 0.20455840229988098, + "learning_rate": 2.112520846198053e-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": 0.008037483179941773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037062566028907896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019781222846359015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022326320968568326, + "weighted_rkd_loss": 0.009890611423179508 + }, + { + "epoch": 3.3500837520938025, + "flow_matching_loss": 0.0228325380012393, + "grad_norm": 0.2642732560634613, + "learning_rate": 2.103451527260074e-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": 0.007869466673582792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033093452220782638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019048278499394656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0228325380012393, + "weighted_rkd_loss": 0.009524139249697328 + }, + { + "epoch": 3.3668341708542715, + "flow_matching_loss": 0.019583611376583578, + "grad_norm": 0.2518172860145569, + "learning_rate": 2.094355811397381e-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": 0.008222047938033938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033477101242169738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0197918058373034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019583611376583578, + "weighted_rkd_loss": 0.0098959029186517 + }, + { + "epoch": 3.3835845896147405, + "flow_matching_loss": 0.024029638990759848, + "grad_norm": 0.19392850995063782, + "learning_rate": 2.0852340964860522e-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": 0.008880104077979923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004434320167638362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022194528207182884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.024029638990759848, + "weighted_rkd_loss": 0.011097264103591442 + }, + { + "epoch": 3.4003350083752095, + "flow_matching_loss": 0.02049247771501541, + "grad_norm": 0.1794871985912323, + "learning_rate": 2.076086781539447e-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": 0.007119336817413569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028993004932999613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01713797440752387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02049247771501541, + "weighted_rkd_loss": 0.008568987203761935 + }, + { + "epoch": 3.4170854271356785, + "flow_matching_loss": 0.017801507376134396, + "grad_norm": 0.20647704601287842, + "learning_rate": 2.066914266690755e-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": 0.008224464533850551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00377469330560416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02022362258285284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017801507376134396, + "weighted_rkd_loss": 0.01011181129142642 + }, + { + "epoch": 3.4338358458961475, + "flow_matching_loss": 0.019913967046886683, + "grad_norm": 0.1910584270954132, + "learning_rate": 2.0577169531754906e-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": 0.008146553998813033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033962710993364455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019689378887414934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019913967046886683, + "weighted_rkd_loss": 0.009844689443707467 + }, + { + "epoch": 3.4505862646566166, + "flow_matching_loss": 0.022410364355891943, + "grad_norm": 0.15049278736114502, + "learning_rate": 2.048495243313942e-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": 0.007043796870857477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003043292067013681, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017130885925143956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.022410364355891943, + "weighted_rkd_loss": 0.008565442962571978 + }, + { + "epoch": 3.4673366834170856, + "flow_matching_loss": 0.019107151962816715, + "grad_norm": 0.1801232546567917, + "learning_rate": 2.039249540493572e-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": 0.006890717707574367, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027747028158046304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016556138079613446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019107151962816715, + "weighted_rkd_loss": 0.008278069039806723 + }, + { + "epoch": 3.4840871021775546, + "flow_matching_loss": 0.01943407179787755, + "grad_norm": 0.21524561941623688, + "learning_rate": 2.0299802491513755e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007901145797222852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037480345461517574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01955032618716359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01943407179787755, + "weighted_rkd_loss": 0.009775163093581795 + }, + { + "epoch": 3.5008375209380236, + "flow_matching_loss": 0.020493100211024285, + "grad_norm": 0.22612208127975464, + "learning_rate": 2.0206877747561828e-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": 0.0077430652920156716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003920659353025257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019406790286302565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020493100211024285, + "weighted_rkd_loss": 0.009703395143151283 + }, + { + "epoch": 3.5175879396984926, + "flow_matching_loss": 0.017947696149349213, + "grad_norm": 0.21178822219371796, + "learning_rate": 2.011372523790927e-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": 0.008696998190134763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004025266319513321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021419262699782848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017947696149349213, + "weighted_rkd_loss": 0.010709631349891424 + }, + { + "epoch": 3.5343383584589616, + "flow_matching_loss": 0.018908549472689627, + "grad_norm": 0.20278148353099823, + "learning_rate": 2.0020349037348643e-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": 0.008784465352073311, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003880465682595968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02144939647987485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018908549472689627, + "weighted_rkd_loss": 0.010724698239937425 + }, + { + "epoch": 3.5510887772194306, + "flow_matching_loss": 0.018494799360632898, + "grad_norm": 0.1947580724954605, + "learning_rate": 1.9926753230457443e-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": 0.009449501568451524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004069603490643203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022968606650829317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018494799360632898, + "weighted_rkd_loss": 0.011484303325414658 + }, + { + "epoch": 3.5678391959798996, + "flow_matching_loss": 0.02321052923798561, + "grad_norm": 0.21995417773723602, + "learning_rate": 1.983294191141948e-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": 0.007246848475188017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003205891884863377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017699589021503925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02321052923798561, + "weighted_rkd_loss": 0.008849794510751963 + }, + { + "epoch": 3.5845896147403686, + "flow_matching_loss": 0.02240783693268895, + "grad_norm": 0.19936057925224304, + "learning_rate": 1.973891918384575e-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": 0.009624690655618906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004620609059929848, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02386999037116766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02240783693268895, + "weighted_rkd_loss": 0.01193499518558383 + }, + { + "epoch": 3.6013400335008376, + "flow_matching_loss": 0.02322760485112667, + "grad_norm": 0.21677754819393158, + "learning_rate": 1.9644689160594945e-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": 0.0078708321088925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003928346477914601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019670010544359684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02322760485112667, + "weighted_rkd_loss": 0.009835005272179842 + }, + { + "epoch": 3.6180904522613067, + "flow_matching_loss": 0.019495286140590908, + "grad_norm": 0.22844098508358002, + "learning_rate": 1.9550255963593553e-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": 0.007603819016367197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035069640493020416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01871460219845176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019495286140590908, + "weighted_rkd_loss": 0.00935730109922588 + }, + { + "epoch": 3.6348408710217757, + "flow_matching_loss": 0.020784722454845906, + "grad_norm": 0.19066329300403595, + "learning_rate": 1.9455623723655524e-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": 0.006900070700794458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033538440940901635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01715398533269763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020784722454845906, + "weighted_rkd_loss": 0.008576992666348815 + }, + { + "epoch": 3.6515912897822447, + "flow_matching_loss": 0.02208636198192835, + "grad_norm": 0.12610961496829987, + "learning_rate": 1.9360796580301606e-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": 0.006914446130394936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028566413442604245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016685533430427314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02208636198192835, + "weighted_rkd_loss": 0.008342766715213657 + }, + { + "epoch": 3.6683417085427137, + "flow_matching_loss": 0.02097361385822296, + "grad_norm": 0.19243593513965607, + "learning_rate": 1.9265778681578245e-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": 0.007681887224316597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037624599412083627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0191262342967093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02097361385822296, + "weighted_rkd_loss": 0.00956311714835465 + }, + { + "epoch": 3.6850921273031827, + "flow_matching_loss": 0.019382779020816088, + "grad_norm": 0.2503208518028259, + "learning_rate": 1.9170574183876143e-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": 0.008374608866870404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038465243531391026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02059574192389846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019382779020816088, + "weighted_rkd_loss": 0.01029787096194923 + }, + { + "epoch": 3.7018425460636517, + "flow_matching_loss": 0.01804492184892297, + "grad_norm": 0.16802993416786194, + "learning_rate": 1.9075187251748452e-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": 0.009310635272413492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043102250900119545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02293149596080184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01804492184892297, + "weighted_rkd_loss": 0.01146574798040092 + }, + { + "epoch": 3.7185929648241207, + "flow_matching_loss": 0.020559789799153804, + "grad_norm": 0.20331931114196777, + "learning_rate": 1.8979622057728596e-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": 0.007381893135607243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030734885251149533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017837275005877017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020559789799153804, + "weighted_rkd_loss": 0.008918637502938509 + }, + { + "epoch": 3.7353433835845897, + "flow_matching_loss": 0.021128038968890907, + "grad_norm": 0.2322036772966385, + "learning_rate": 1.8883882782147748e-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": 0.007587933773174882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003291653376072645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01846752082929015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021128038968890907, + "weighted_rkd_loss": 0.009233760414645076 + }, + { + "epoch": 3.7520938023450587, + "flow_matching_loss": 0.020190212596207856, + "grad_norm": 0.2527245879173279, + "learning_rate": 1.8787973612951967e-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": 0.008083526277914643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038183052791282536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01998535804450512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020190212596207856, + "weighted_rkd_loss": 0.00999267902225256 + }, + { + "epoch": 3.7688442211055277, + "flow_matching_loss": 0.020125599578022957, + "grad_norm": 0.18612495064735413, + "learning_rate": 1.8691898745519028e-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": 0.007300873147323727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002893792022950947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017495538201183082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020125599578022957, + "weighted_rkd_loss": 0.008747769100591541 + }, + { + "epoch": 3.7855946398659968, + "flow_matching_loss": 0.02288515316322446, + "grad_norm": 0.2038274109363556, + "learning_rate": 1.8595662382474863e-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": 0.007445386983454228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003284679213538766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01817545313388109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02288515316322446, + "weighted_rkd_loss": 0.009087726566940546 + }, + { + "epoch": 3.8023450586264658, + "flow_matching_loss": 0.01813176032155752, + "grad_norm": 0.15173381567001343, + "learning_rate": 1.8499268733509766e-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": 0.007210503658279776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003472392586991191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01789340013638139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01813176032155752, + "weighted_rkd_loss": 0.008946700068190695 + }, + { + "epoch": 3.819095477386935, + "flow_matching_loss": 0.021203831862658264, + "grad_norm": 0.21146638691425323, + "learning_rate": 1.84027220151942e-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": 0.0065764828585088255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003243767865933478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016396733839064835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.021203831862658264, + "weighted_rkd_loss": 0.008198366919532418 + }, + { + "epoch": 3.835845896147404, + "flow_matching_loss": 0.02068155473098159, + "grad_norm": 0.22891871631145477, + "learning_rate": 1.83060264507944e-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": 0.008209372824057936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003933777997735888, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02035252321511507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02068155473098159, + "weighted_rkd_loss": 0.010176261607557534 + }, + { + "epoch": 3.852596314907873, + "flow_matching_loss": 0.020051562413573264, + "grad_norm": 0.14797094464302063, + "learning_rate": 1.82091862700876e-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": 0.00664952090010047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002952263760380447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01625130567699671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020051562413573264, + "weighted_rkd_loss": 0.008125652838498355 + }, + { + "epoch": 3.869346733668342, + "flow_matching_loss": 0.02093609981238842, + "grad_norm": 0.2215096801519394, + "learning_rate": 1.8112205709177023e-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": 0.00714435912668705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003198638209141791, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017487356252968312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02093609981238842, + "weighted_rkd_loss": 0.008743678126484156 + }, + { + "epoch": 3.886097152428811, + "flow_matching_loss": 0.019973818585276602, + "grad_norm": 0.18120259046554565, + "learning_rate": 1.8015089010306585e-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": 0.0070549966767430305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003187867905944586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01729786125943065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019973818585276602, + "weighted_rkd_loss": 0.008648930629715324 + }, + { + "epoch": 3.90284757118928, + "flow_matching_loss": 0.019720715191215276, + "grad_norm": 0.22681787610054016, + "learning_rate": 1.7917840421675298e-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": 0.007141135260462761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00324872734490782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017530998028814793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019720715191215276, + "weighted_rkd_loss": 0.008765499014407397 + }, + { + "epoch": 3.919597989949749, + "flow_matching_loss": 0.018794780503958462, + "grad_norm": 0.21289195120334625, + "learning_rate": 1.7820464197251474e-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": 0.006023052660748362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029200530145317314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014966158475726844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018794780503958462, + "weighted_rkd_loss": 0.007483079237863422 + }, + { + "epoch": 3.936348408710218, + "flow_matching_loss": 0.02315736413002014, + "grad_norm": 0.22471879422664642, + "learning_rate": 1.7722964596586627e-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": 0.006645664293318987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028370591229759158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01612838786095381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02315736413002014, + "weighted_rkd_loss": 0.008064193930476905 + }, + { + "epoch": 3.953098827470687, + "flow_matching_loss": 0.019443149212747812, + "grad_norm": 0.19242846965789795, + "learning_rate": 1.7625345884629144e-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": 0.007972657866775989, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036050250404514373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019550340808928014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019443149212747812, + "weighted_rkd_loss": 0.009775170404464007 + }, + { + "epoch": 3.969849246231156, + "flow_matching_loss": 0.02088850997388363, + "grad_norm": 0.1732814460992813, + "learning_rate": 1.7527612331537722e-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": 0.006496066693216562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002922359528020024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015914492681622504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02088850997388363, + "weighted_rkd_loss": 0.007957246340811252 + }, + { + "epoch": 3.986599664991625, + "flow_matching_loss": 0.019626106694340707, + "grad_norm": 0.1980288326740265, + "learning_rate": 1.7429768212494593e-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": 0.006757032917812466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003019305528141558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016533371340483428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019626106694340707, + "weighted_rkd_loss": 0.008266685670241714 + }, + { + "epoch": 4.0033500837520934, + "flow_matching_loss": 0.01870811516419053, + "grad_norm": 0.25888529419898987, + "learning_rate": 1.7331817807518497e-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": 0.008479700377210974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036272499710321426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020586650539189576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01870811516419053, + "weighted_rkd_loss": 0.010293325269594788 + }, + { + "epoch": 4.0201005025125625, + "flow_matching_loss": 0.019916127249598504, + "grad_norm": 0.18142595887184143, + "learning_rate": 1.723376540127745e-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": 0.006715080700814724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028278434183448553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016258004866540433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019916127249598504, + "weighted_rkd_loss": 0.008129002433270216 + }, + { + "epoch": 4.0368509212730315, + "flow_matching_loss": 0.01878450671210885, + "grad_norm": 0.23574140667915344, + "learning_rate": 1.7135615282901357e-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": 0.0064266437198966745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003169252898078412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01602254034951329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01878450671210885, + "weighted_rkd_loss": 0.008011270174756646 + }, + { + "epoch": 4.0536013400335005, + "flow_matching_loss": 0.020211434923112392, + "grad_norm": 0.20220205187797546, + "learning_rate": 1.7037371745794358e-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": 0.006330230925232172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002878256363328546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01553871827200055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020211434923112392, + "weighted_rkd_loss": 0.007769359136000275 + }, + { + "epoch": 4.0703517587939695, + "flow_matching_loss": 0.019851094391196967, + "grad_norm": 0.2334161400794983, + "learning_rate": 1.6939039087447026e-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": 0.0066378221847116945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003039022069424391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016314666345715524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019851094391196967, + "weighted_rkd_loss": 0.008157333172857762 + }, + { + "epoch": 4.0871021775544385, + "flow_matching_loss": 0.020226925425231457, + "grad_norm": 0.1534281224012375, + "learning_rate": 1.6840621609248397e-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": 0.007128366222605109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003230903437361121, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017487635742872952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020226925425231457, + "weighted_rkd_loss": 0.008743817871436476 + }, + { + "epoch": 4.1038525963149075, + "flow_matching_loss": 0.019695611950010062, + "grad_norm": 0.2591371536254883, + "learning_rate": 1.6742123616297794e-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": 0.007243454223498702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00313732223585248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017624230589717627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019695611950010062, + "weighted_rkd_loss": 0.008812115294858813 + }, + { + "epoch": 4.1206030150753765, + "flow_matching_loss": 0.018704015389084817, + "grad_norm": 0.13702625036239624, + "learning_rate": 1.664354941721651e-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": 0.007517511304467916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003035618958529085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018070641439408064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018704015389084817, + "weighted_rkd_loss": 0.009035320719704032 + }, + { + "epoch": 4.1373534338358455, + "flow_matching_loss": 0.017411495372653008, + "grad_norm": 0.19919848442077637, + "learning_rate": 1.6544903323959355e-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": 0.007433804357424379, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003234395873732865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01810200437903404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017411495372653008, + "weighted_rkd_loss": 0.00905100218951702 + }, + { + "epoch": 4.1541038525963145, + "flow_matching_loss": 0.02045935718342662, + "grad_norm": 0.21578043699264526, + "learning_rate": 1.6446189651626023e-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": 0.007500055991113186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003137087286449969, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01813719905912876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02045935718342662, + "weighted_rkd_loss": 0.00906859952956438 + }, + { + "epoch": 4.1708542713567835, + "flow_matching_loss": 0.02151382314041257, + "grad_norm": 0.16690842807292938, + "learning_rate": 1.6347412718272316e-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": 0.0064533458091318606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029357226798310877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01584241455420852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02151382314041257, + "weighted_rkd_loss": 0.00792120727710426 + }, + { + "epoch": 4.187604690117253, + "flow_matching_loss": 0.016593755688518286, + "grad_norm": 0.1668848693370819, + "learning_rate": 1.624857684472131e-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": 0.006326651386916638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00287430175812915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015527604427188636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016593755688518286, + "weighted_rkd_loss": 0.007763802213594318 + }, + { + "epoch": 4.204355108877722, + "flow_matching_loss": 0.018163218908011914, + "grad_norm": 0.21210342645645142, + "learning_rate": 1.6149686354374283e-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": 0.006518031936138868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030372726498171685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016073336359113455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018163218908011914, + "weighted_rkd_loss": 0.008036668179556727 + }, + { + "epoch": 4.221105527638191, + "flow_matching_loss": 0.01736935544759035, + "grad_norm": 0.18703977763652802, + "learning_rate": 1.605074557302165e-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": 0.007781515922397375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003655482176691294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01921851420775056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01736935544759035, + "weighted_rkd_loss": 0.00960925710387528 + }, + { + "epoch": 4.23785594639866, + "flow_matching_loss": 0.02085604052990675, + "grad_norm": 0.1353197544813156, + "learning_rate": 1.59517588286537e-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": 0.007296613347716629, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029753129929304123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0175685397349298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02085604052990675, + "weighted_rkd_loss": 0.0087842698674649 + }, + { + "epoch": 4.254606365159129, + "flow_matching_loss": 0.020376010704785584, + "grad_norm": 0.1928730607032776, + "learning_rate": 1.5852730451271324e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007122312113642693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003080862946808338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017325487174093723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020376010704785584, + "weighted_rkd_loss": 0.008662743587046861 + }, + { + "epoch": 4.271356783919598, + "flow_matching_loss": 0.018556658551096915, + "grad_norm": 0.18790344893932343, + "learning_rate": 1.5753664772696546e-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": 0.006062540132552386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028738549328409134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014998935256153345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018556658551096915, + "weighted_rkd_loss": 0.007499467628076672 + }, + { + "epoch": 4.288107202680067, + "flow_matching_loss": 0.018578741513192654, + "grad_norm": 0.20445463061332703, + "learning_rate": 1.565456612638307e-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": 0.006489685783162713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027945277863182126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015773899294435977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018578741513192654, + "weighted_rkd_loss": 0.007886949647217988 + }, + { + "epoch": 4.304857621440536, + "flow_matching_loss": 0.02046471843495965, + "grad_norm": 0.1576392948627472, + "learning_rate": 1.5555438847226732e-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": 0.006184978596866131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025121036567725243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014882060885429382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02046471843495965, + "weighted_rkd_loss": 0.007441030442714691 + }, + { + "epoch": 4.321608040201005, + "flow_matching_loss": 0.02033666577190161, + "grad_norm": 0.19884538650512695, + "learning_rate": 1.545628727137584e-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": 0.006643129396252334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029675282421521842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016253787092864513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02033666577190161, + "weighted_rkd_loss": 0.008126893546432257 + }, + { + "epoch": 4.338358458961474, + "flow_matching_loss": 0.01932725394144654, + "grad_norm": 0.1970246136188507, + "learning_rate": 1.535711573604153e-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": 0.00804528738372028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036893374170176687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019779912009835244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01932725394144654, + "weighted_rkd_loss": 0.009889956004917622 + }, + { + "epoch": 4.355108877721943, + "flow_matching_loss": 0.01995522724464536, + "grad_norm": 0.1644182652235031, + "learning_rate": 1.5257928579308032e-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": 0.006598327262327075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033264036406762897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016523058246821164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01995522724464536, + "weighted_rkd_loss": 0.008261529123410582 + }, + { + "epoch": 4.371859296482412, + "flow_matching_loss": 0.0177932090125978, + "grad_norm": 0.21884872019290924, + "learning_rate": 1.5158730139942902e-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": 0.008010930195450783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033667325507849453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019388592801988125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0177932090125978, + "weighted_rkd_loss": 0.009694296400994062 + }, + { + "epoch": 4.388609715242881, + "flow_matching_loss": 0.019758843723684548, + "grad_norm": 0.1710134893655777, + "learning_rate": 1.5059524757207239e-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": 0.00648601446300745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029278623638674616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015899891406297682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019758843723684548, + "weighted_rkd_loss": 0.007949945703148841 + }, + { + "epoch": 4.40536013400335, + "flow_matching_loss": 0.017516419384628536, + "grad_norm": 0.15637044608592987, + "learning_rate": 1.4960316770665866e-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": 0.006116408109664917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025824596639722587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014815275836735965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017516419384628536, + "weighted_rkd_loss": 0.007407637918367982 + }, + { + "epoch": 4.422110552763819, + "flow_matching_loss": 0.019074277114123107, + "grad_norm": 0.22107526659965515, + "learning_rate": 1.4861110519997495e-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": 0.006211187061853707, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028650935622863472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015287467464804649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019074277114123107, + "weighted_rkd_loss": 0.007643733732402324 + }, + { + "epoch": 4.438860971524288, + "flow_matching_loss": 0.01782265156507492, + "grad_norm": 0.23777632415294647, + "learning_rate": 1.4761910344804913e-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": 0.007094258349388838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028426593402400614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01703117610886693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01782265156507492, + "weighted_rkd_loss": 0.008515588054433466 + }, + { + "epoch": 4.455611390284757, + "flow_matching_loss": 0.020024440810084344, + "grad_norm": 0.1772475689649582, + "learning_rate": 1.466272058442516e-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": 0.006881606904789806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002932384598534554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016695598419755698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020024440810084344, + "weighted_rkd_loss": 0.008347799209877849 + }, + { + "epoch": 4.472361809045226, + "flow_matching_loss": 0.01836460642516613, + "grad_norm": 0.1673913151025772, + "learning_rate": 1.4563545577739676e-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": 0.0067969394847750665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003281451168004423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016875330172479153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01836460642516613, + "weighted_rkd_loss": 0.008437665086239577 + }, + { + "epoch": 4.489112227805695, + "flow_matching_loss": 0.016627813410013913, + "grad_norm": 0.16218630969524384, + "learning_rate": 1.4464389662984543e-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": 0.006768457312136889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003101827809587121, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01663874238729477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016627813410013913, + "weighted_rkd_loss": 0.008319371193647385 + }, + { + "epoch": 4.505862646566165, + "flow_matching_loss": 0.016898567136377097, + "grad_norm": 0.14560790359973907, + "learning_rate": 1.4365257177560684e-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": 0.005781614547595381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002422610018402338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013985839113593102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016898567136377097, + "weighted_rkd_loss": 0.006992919556796551 + }, + { + "epoch": 4.522613065326633, + "flow_matching_loss": 0.01971247997134924, + "grad_norm": 0.14987064898014069, + "learning_rate": 1.426615245784416e-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": 0.007260903157293797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002818218374159187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01734002446755767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01971247997134924, + "weighted_rkd_loss": 0.008670012233778834 + }, + { + "epoch": 4.539363484087103, + "flow_matching_loss": 0.017527775652706624, + "grad_norm": 0.1931697577238083, + "learning_rate": 1.4167079838996468e-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": 0.007004645606502891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003280059481039643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017289350647479294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017527775652706624, + "weighted_rkd_loss": 0.008644675323739647 + }, + { + "epoch": 4.556113902847571, + "flow_matching_loss": 0.018044522684067487, + "grad_norm": 0.18841363489627838, + "learning_rate": 1.4068043654774897e-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": 0.0072848881827667356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003721295588184148, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0182910718023777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018044522684067487, + "weighted_rkd_loss": 0.00914553590118885 + }, + { + "epoch": 4.572864321608041, + "flow_matching_loss": 0.016927159391343594, + "grad_norm": 0.18175353109836578, + "learning_rate": 1.3969048237342992e-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": 0.006235395558178425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026640282245352863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015134819317609072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016927159391343594, + "weighted_rkd_loss": 0.007567409658804536 + }, + { + "epoch": 4.589614740368509, + "flow_matching_loss": 0.020692834444344044, + "grad_norm": 0.17616070806980133, + "learning_rate": 1.3870097917081011e-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": 0.00688162581063807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029462887556292116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01670954031869769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020692834444344044, + "weighted_rkd_loss": 0.008354770159348846 + }, + { + "epoch": 4.606365159128979, + "flow_matching_loss": 0.01656912127509713, + "grad_norm": 0.17389723658561707, + "learning_rate": 1.3771197022396508e-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": 0.0060537876561284065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002805469243321568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014913044683635236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01656912127509713, + "weighted_rkd_loss": 0.007456522341817618 + }, + { + "epoch": 4.623115577889447, + "flow_matching_loss": 0.018882535118609666, + "grad_norm": 0.1804601401090622, + "learning_rate": 1.3672349879535015e-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": 0.0065174739807844165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026487289811484515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015683677047491074, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018882535118609666, + "weighted_rkd_loss": 0.007841838523745537 + }, + { + "epoch": 4.639865996649917, + "flow_matching_loss": 0.018018264696002005, + "grad_norm": 0.18113377690315247, + "learning_rate": 1.3573560812390781e-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": 0.006505385902710259, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00266862609423697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015679398085922003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018018264696002005, + "weighted_rkd_loss": 0.007839699042961001 + }, + { + "epoch": 4.656616415410385, + "flow_matching_loss": 0.017154605966061352, + "grad_norm": 0.22344084084033966, + "learning_rate": 1.3474834142317635e-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": 0.00781757920049131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003766122250817716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019401280861347913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017154605966061352, + "weighted_rkd_loss": 0.009700640430673957 + }, + { + "epoch": 4.673366834170855, + "flow_matching_loss": 0.020884703472256662, + "grad_norm": 0.16884377598762512, + "learning_rate": 1.3376174187939955e-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": 0.00663802158087492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002596668968908489, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015872712060809136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020884703472256662, + "weighted_rkd_loss": 0.007936356030404568 + }, + { + "epoch": 4.690117252931323, + "flow_matching_loss": 0.01987990001216531, + "grad_norm": 0.17503158748149872, + "learning_rate": 1.3277585264963761e-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": 0.006741104461252689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00272078113630414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01620299005880952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01987990001216531, + "weighted_rkd_loss": 0.00810149502940476 + }, + { + "epoch": 4.706867671691793, + "flow_matching_loss": 0.018524412903934716, + "grad_norm": 0.2198275923728943, + "learning_rate": 1.3179071685987928e-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": 0.006492939265444875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026159399072639643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015601818356662989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018524412903934716, + "weighted_rkd_loss": 0.0078009091783314945 + }, + { + "epoch": 4.723618090452261, + "flow_matching_loss": 0.018392641749233007, + "grad_norm": 0.15910054743289948, + "learning_rate": 1.308063776031554e-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": 0.006039766082540155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022717000334523617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01435123234987259, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018392641749233007, + "weighted_rkd_loss": 0.007175616174936295 + }, + { + "epoch": 4.740368509212731, + "flow_matching_loss": 0.016925379633903503, + "grad_norm": 0.1666567623615265, + "learning_rate": 1.2982287793765389e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005827722325921059, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002627475606277585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014282920211553574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016925379633903503, + "weighted_rkd_loss": 0.007141460105776787 + }, + { + "epoch": 4.757118927973199, + "flow_matching_loss": 0.016931863874197005, + "grad_norm": 0.16195693612098694, + "learning_rate": 1.2884026088483616e-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": 0.006095975963398814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00292929713614285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015121249109506607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016931863874197005, + "weighted_rkd_loss": 0.007560624554753303 + }, + { + "epoch": 4.773869346733669, + "flow_matching_loss": 0.02151737492531538, + "grad_norm": 0.16743698716163635, + "learning_rate": 1.2785856942755543e-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": 0.006752799730747938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033896766020916403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016895275842398406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.02151737492531538, + "weighted_rkd_loss": 0.008447637921199203 + }, + { + "epoch": 4.790619765494137, + "flow_matching_loss": 0.018972883094102146, + "grad_norm": 0.21053996682167053, + "learning_rate": 1.2687784650817616e-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": 0.006391771929338574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025758426752872767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015359386708587408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018972883094102146, + "weighted_rkd_loss": 0.007679693354293704 + }, + { + "epoch": 4.807370184254607, + "flow_matching_loss": 0.020155242085456847, + "grad_norm": 0.15815846621990204, + "learning_rate": 1.2589813502669603e-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": 0.0053243459668010475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020828464068472384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012731538340449332, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020155242085456847, + "weighted_rkd_loss": 0.006365769170224666 + }, + { + "epoch": 4.824120603015075, + "flow_matching_loss": 0.019101917929947376, + "grad_norm": 0.14777660369873047, + "learning_rate": 1.2491947783886907e-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": 0.006484691519290209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031254583853296937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016094841342419387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019101917929947376, + "weighted_rkd_loss": 0.008047420671209693 + }, + { + "epoch": 4.840871021775545, + "flow_matching_loss": 0.01791600426658988, + "grad_norm": 0.12246904522180557, + "learning_rate": 1.2394191775433106e-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": 0.005786749953404069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028565174899995325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014430017489939929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01791600426658988, + "weighted_rkd_loss": 0.007215008744969964 + }, + { + "epoch": 4.857621440536013, + "flow_matching_loss": 0.016301574185490608, + "grad_norm": 0.1843765676021576, + "learning_rate": 1.2296549753472696e-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": 0.005566553748212755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026679943315684796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013801102060824633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016301574185490608, + "weighted_rkd_loss": 0.006900551030412316 + }, + { + "epoch": 4.874371859296483, + "flow_matching_loss": 0.01598440548405051, + "grad_norm": 0.17817851901054382, + "learning_rate": 1.2199025989184032e-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": 0.00497903733048588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021776941372081637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012135768868029117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01598440548405051, + "weighted_rkd_loss": 0.006067884434014559 + }, + { + "epoch": 4.891122278056951, + "flow_matching_loss": 0.018774052895605565, + "grad_norm": 0.13188794255256653, + "learning_rate": 1.21016247485725e-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": 0.005346168461255729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025814786669798194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013273815717548132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018774052895605565, + "weighted_rkd_loss": 0.006636907858774066 + }, + { + "epoch": 4.907872696817421, + "flow_matching_loss": 0.018068937305361032, + "grad_norm": 0.1632096916437149, + "learning_rate": 1.2004350292283897e-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": 0.005146085587330163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023135139606893064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012605685275048018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018068937305361032, + "weighted_rkd_loss": 0.006302842637524009 + }, + { + "epoch": 4.924623115577889, + "flow_matching_loss": 0.017883090209215878, + "grad_norm": 0.15789315104484558, + "learning_rate": 1.1907206875418068e-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": 0.006365909287706017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027634695288725197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01549528781324625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017883090209215878, + "weighted_rkd_loss": 0.007747643906623125 + }, + { + "epoch": 4.941373534338359, + "flow_matching_loss": 0.018130379170179366, + "grad_norm": 0.18408702313899994, + "learning_rate": 1.1810198747342769e-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": 0.0056016219314187765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025636515463702382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01376689551398158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018130379170179366, + "weighted_rkd_loss": 0.00688344775699079 + }, + { + "epoch": 4.958123953098827, + "flow_matching_loss": 0.015793959517031907, + "grad_norm": 0.1873013824224472, + "learning_rate": 1.1713330151507786e-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": 0.006022406509146094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025855567771941425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014630369748920203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015793959517031907, + "weighted_rkd_loss": 0.007315184874460101 + }, + { + "epoch": 4.974874371859297, + "flow_matching_loss": 0.01843378497287631, + "grad_norm": 0.1450514942407608, + "learning_rate": 1.161660532525931e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063908693846315145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002764756348915398, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015546495281159877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01843378497287631, + "weighted_rkd_loss": 0.0077732476405799385 + }, + { + "epoch": 4.991624790619765, + "flow_matching_loss": 0.015466341562569142, + "grad_norm": 0.17179778218269348, + "learning_rate": 1.1520028499654584e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060694361571222545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024145430303178727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014553415589034557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015466341562569142, + "weighted_rkd_loss": 0.007276707794517279 + }, + { + "epoch": 5.008375209380235, + "flow_matching_loss": 0.017944001685827972, + "grad_norm": 0.21506550908088684, + "learning_rate": 1.1423603899276833e-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": 0.006832077959552407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002628091024234891, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016292246989905834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017944001685827972, + "weighted_rkd_loss": 0.008146123494952917 + }, + { + "epoch": 5.025125628140704, + "flow_matching_loss": 0.017549295723438264, + "grad_norm": 0.26095050573349, + "learning_rate": 1.1327335742050442e-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": 0.006050047976896167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002698522701393813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01479861866682768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017549295723438264, + "weighted_rkd_loss": 0.00739930933341384 + }, + { + "epoch": 5.041876046901173, + "flow_matching_loss": 0.016273344960063697, + "grad_norm": 0.1535094678401947, + "learning_rate": 1.1231228239056467e-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": 0.005975645314902067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023796642548404635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014330955035984517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016273344960063697, + "weighted_rkd_loss": 0.007165477517992258 + }, + { + "epoch": 5.058626465661642, + "flow_matching_loss": 0.018688787426799535, + "grad_norm": 0.1590380221605301, + "learning_rate": 1.113528559434845e-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": 0.005144167877733708, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002109654329251498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012397990189492702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018688787426799535, + "weighted_rkd_loss": 0.006198995094746351 + }, + { + "epoch": 5.075376884422111, + "flow_matching_loss": 0.020107445027679204, + "grad_norm": 0.2060288041830063, + "learning_rate": 1.1039512004768476e-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": 0.006088935793377459, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002645562624093145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014823434315621853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020107445027679204, + "weighted_rkd_loss": 0.007411717157810927 + }, + { + "epoch": 5.09212730318258, + "flow_matching_loss": 0.017590080946683885, + "grad_norm": 0.1265738606452942, + "learning_rate": 1.0943911659763612e-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": 0.006028936221264302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025103176943957807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014568190462887287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017590080946683885, + "weighted_rkd_loss": 0.007284095231443644 + }, + { + "epoch": 5.108877721943049, + "flow_matching_loss": 0.017241567280143498, + "grad_norm": 0.15954196453094482, + "learning_rate": 1.084848874120266e-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": 0.006110600777901709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026879373588599266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01490913899615407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017241567280143498, + "weighted_rkd_loss": 0.007454569498077035 + }, + { + "epoch": 5.125628140703517, + "flow_matching_loss": 0.018878405541181566, + "grad_norm": 0.17765270173549652, + "learning_rate": 1.0753247423193196e-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": 0.00597507394850254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027689284761436284, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014719076454639435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018878405541181566, + "weighted_rkd_loss": 0.007359538227319717 + }, + { + "epoch": 5.142378559463987, + "flow_matching_loss": 0.017817400768399237, + "grad_norm": 0.17209725081920624, + "learning_rate": 1.0658191871899012e-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": 0.005254556564614177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024205024936236443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012929615657776594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017817400768399237, + "weighted_rkd_loss": 0.006464807828888297 + }, + { + "epoch": 5.159128978224456, + "flow_matching_loss": 0.01816714182496071, + "grad_norm": 0.14104831218719482, + "learning_rate": 1.0563326245357852e-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": 0.0050902100512757896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002153707609977573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012334127770736813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01816714182496071, + "weighted_rkd_loss": 0.006167063885368407 + }, + { + "epoch": 5.175879396984925, + "flow_matching_loss": 0.01670421687886119, + "grad_norm": 0.1472315937280655, + "learning_rate": 1.0468654693299557e-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": 0.0054107870440930125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027880519977770747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013609626144170762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01670421687886119, + "weighted_rkd_loss": 0.006804813072085381 + }, + { + "epoch": 5.192629815745394, + "flow_matching_loss": 0.015585028752684593, + "grad_norm": 0.24463646113872528, + "learning_rate": 1.0374181356964501e-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": 0.006263877777382731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026189771597273648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015146732609719039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015585028752684593, + "weighted_rkd_loss": 0.007573366304859519 + }, + { + "epoch": 5.209380234505863, + "flow_matching_loss": 0.016866875905543567, + "grad_norm": 0.20484580099582672, + "learning_rate": 1.0279910368922465e-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": 0.005633864691480995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025424073450267316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013810136541724204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016866875905543567, + "weighted_rkd_loss": 0.006905068270862102 + }, + { + "epoch": 5.226130653266332, + "flow_matching_loss": 0.015790032222867012, + "grad_norm": 0.1716005951166153, + "learning_rate": 1.018584585289186e-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": 0.005199952656403184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00229923086008057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012699136231094598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015790032222867012, + "weighted_rkd_loss": 0.006349568115547299 + }, + { + "epoch": 5.242881072026801, + "flow_matching_loss": 0.017556129675358533, + "grad_norm": 0.22674204409122467, + "learning_rate": 1.0091991923559348e-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": 0.005729359481483698, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024778502178378404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013936569122597576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017556129675358533, + "weighted_rkd_loss": 0.006968284561298788 + }, + { + "epoch": 5.259631490787269, + "flow_matching_loss": 0.017610132601112126, + "grad_norm": 0.14236882328987122, + "learning_rate": 9.99835268639984e-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": 0.00562389271799475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023109326721169055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01355871818959713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017610132601112126, + "weighted_rkd_loss": 0.006779359094798565 + }, + { + "epoch": 5.276381909547739, + "flow_matching_loss": 0.015706593077629803, + "grad_norm": 0.1687908172607422, + "learning_rate": 9.904932237496913e-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": 0.00528958635404706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022576195537112654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012836792226880788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015706593077629803, + "weighted_rkd_loss": 0.006418396113440394 + }, + { + "epoch": 5.293132328308207, + "flow_matching_loss": 0.014574210066348315, + "grad_norm": 0.16634498536586761, + "learning_rate": 9.811734663363659e-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": 0.005389407277107239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002465158875565976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013243973534554244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014574210066348315, + "weighted_rkd_loss": 0.006621986767277122 + }, + { + "epoch": 5.309882747068677, + "flow_matching_loss": 0.01800208482891321, + "grad_norm": 0.197093203663826, + "learning_rate": 9.718764040763875e-06, + "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": 0.006899569206871093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002856101421639323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01665523974224925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01800208482891321, + "weighted_rkd_loss": 0.008327619871124625 + }, + { + "epoch": 5.326633165829146, + "flow_matching_loss": 0.015227837394922972, + "grad_norm": 0.18592070043087006, + "learning_rate": 9.626024436533777e-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": 0.006064104102551937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002545801061205566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014674009475857019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015227837394922972, + "weighted_rkd_loss": 0.007337004737928509 + }, + { + "epoch": 5.343383584589615, + "flow_matching_loss": 0.018127095606178045, + "grad_norm": 0.17893438041210175, + "learning_rate": 9.533519907404095e-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": 0.005131047265604138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023227728554047642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012584867607802152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018127095606178045, + "weighted_rkd_loss": 0.006292433803901076 + }, + { + "epoch": 5.360134003350084, + "flow_matching_loss": 0.01897403774783015, + "grad_norm": 0.1659892350435257, + "learning_rate": 9.4412544998226e-06, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005119132879190147, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022993266000412406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012537592463195324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01897403774783015, + "weighted_rkd_loss": 0.006268796231597662 + }, + { + "epoch": 5.376884422110553, + "flow_matching_loss": 0.017207826394587754, + "grad_norm": 0.15151554346084595, + "learning_rate": 9.349232249777124e-06, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061891798861324785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025767922634258866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014955152198672295, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017207826394587754, + "weighted_rkd_loss": 0.0074775760993361475 + }, + { + "epoch": 5.393634840871022, + "flow_matching_loss": 0.018857961520552635, + "grad_norm": 0.2096855342388153, + "learning_rate": 9.257457182618985e-06, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005416905065067113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002632979757618159, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01346679013222456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018857961520552635, + "weighted_rkd_loss": 0.00673339506611228 + }, + { + "epoch": 5.410385259631491, + "flow_matching_loss": 0.01634107967838645, + "grad_norm": 0.20736907422542572, + "learning_rate": 9.165933312886944e-06, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065269402228295805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027695797849446534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015823460090905427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01634107967838645, + "weighted_rkd_loss": 0.007911730045452713 + }, + { + "epoch": 5.42713567839196, + "flow_matching_loss": 0.016097908932715656, + "grad_norm": 0.15641118586063385, + "learning_rate": 9.07466464413155e-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": 0.0045119600370526316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002176347211934626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011200267216190695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016097908932715656, + "weighted_rkd_loss": 0.005600133608095348 + }, + { + "epoch": 5.443886097152429, + "flow_matching_loss": 0.018399234488606454, + "grad_norm": 0.20723454654216766, + "learning_rate": 8.983655168740048e-06, + "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": 0.005364138376899063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022491870215162633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01297746361233294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018399234488606454, + "weighted_rkd_loss": 0.00648873180616647 + }, + { + "epoch": 5.460636515912898, + "flow_matching_loss": 0.016172559838742016, + "grad_norm": 0.17914651334285736, + "learning_rate": 8.892908867761718e-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": 0.004746598890051246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020948843797668816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01158808209002018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016172559838742016, + "weighted_rkd_loss": 0.00579404104501009 + }, + { + "epoch": 5.477386934673367, + "flow_matching_loss": 0.01892400048673153, + "grad_norm": 0.1844613254070282, + "learning_rate": 8.802429710733745e-06, + "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": 0.004971841280348599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020754300872795283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012019112706184387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01892400048673153, + "weighted_rkd_loss": 0.006009556353092194 + }, + { + "epoch": 5.494137353433836, + "flow_matching_loss": 0.017203427013009788, + "grad_norm": 0.14044742286205292, + "learning_rate": 8.71222165550757e-06, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005311322677880525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002102222468238324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012724867742508649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017203427013009788, + "weighted_rkd_loss": 0.0063624338712543246 + }, + { + "epoch": 5.510887772194305, + "flow_matching_loss": 0.017448701057583095, + "grad_norm": 0.13229121267795563, + "learning_rate": 8.622288648075748e-06, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004944969573989511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022162436041980984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01210618270561099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017448701057583095, + "weighted_rkd_loss": 0.006053091352805495 + }, + { + "epoch": 5.527638190954773, + "flow_matching_loss": 0.018544258456677197, + "grad_norm": 0.18051110208034515, + "learning_rate": 8.532634622399391e-06, + "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": 0.005711289774626494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021691814879886808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013591760955750942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018544258456677197, + "weighted_rkd_loss": 0.006795880477875471 + }, + { + "epoch": 5.544388609715243, + "flow_matching_loss": 0.01410124646499753, + "grad_norm": 0.11878155916929245, + "learning_rate": 8.44326350023601e-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": 0.004789017117582261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019207892357371748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011498823389410972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01410124646499753, + "weighted_rkd_loss": 0.005749411694705486 + }, + { + "epoch": 5.561139028475711, + "flow_matching_loss": 0.016507032420486213, + "grad_norm": 0.1763717085123062, + "learning_rate": 8.354179190968002e-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": 0.005711464490741491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002308214281219989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01373114325106144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016507032420486213, + "weighted_rkd_loss": 0.00686557162553072 + }, + { + "epoch": 5.577889447236181, + "flow_matching_loss": 0.020039274729788302, + "grad_norm": 0.13657470047473907, + "learning_rate": 8.265385591431661e-06, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005672144400887191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025627062539570035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013906995020806789, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.020039274729788302, + "weighted_rkd_loss": 0.006953497510403394 + }, + { + "epoch": 5.594639865996649, + "flow_matching_loss": 0.015484216436743737, + "grad_norm": 0.1777724176645279, + "learning_rate": 8.176886585746671e-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": 0.006078959535807371, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025322111207060514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014690130297094584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015484216436743737, + "weighted_rkd_loss": 0.007345065148547292 + }, + { + "epoch": 5.611390284757119, + "flow_matching_loss": 0.01517918286845088, + "grad_norm": 0.16232475638389587, + "learning_rate": 8.088686045146247e-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": 0.004736245633102954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024781554588116706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011950646806508303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01517918286845088, + "weighted_rkd_loss": 0.005975323403254152 + }, + { + "epoch": 5.628140703517588, + "flow_matching_loss": 0.015617287997156382, + "grad_norm": 0.160396009683609, + "learning_rate": 8.00078782780775e-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": 0.004729195917025209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002444211218971759, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011902603134512902, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015617287997156382, + "weighted_rkd_loss": 0.005951301567256451 + }, + { + "epoch": 5.644891122278057, + "flow_matching_loss": 0.019811277277767658, + "grad_norm": 0.1667182594537735, + "learning_rate": 7.913195778683979e-06, + "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": 0.004932631156407297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002067512331996113, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011932774633169174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019811277277767658, + "weighted_rkd_loss": 0.005966387316584587 + }, + { + "epoch": 5.661641541038526, + "flow_matching_loss": 0.016531050205230713, + "grad_norm": 0.15772488713264465, + "learning_rate": 7.825913729334908e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00438170931302011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018925598124042153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0106559784617275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016531050205230713, + "weighted_rkd_loss": 0.00532798923086375 + }, + { + "epoch": 5.678391959798995, + "flow_matching_loss": 0.018603511899709702, + "grad_norm": 0.1836947351694107, + "learning_rate": 7.73894549776013e-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": 0.004417502367869019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019307143054902553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01076571922749281, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018603511899709702, + "weighted_rkd_loss": 0.005382859613746405 + }, + { + "epoch": 5.695142378559464, + "flow_matching_loss": 0.0157127789221704, + "grad_norm": 0.15395602583885193, + "learning_rate": 7.65229488823184e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005178571795113385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020820079836994408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012439151760190725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0157127789221704, + "weighted_rkd_loss": 0.006219575880095363 + }, + { + "epoch": 5.711892797319933, + "flow_matching_loss": 0.01738949567079544, + "grad_norm": 0.22008782625198364, + "learning_rate": 7.565965691128396e-06, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005835405667312443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022089136065915225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01387972510419786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01738949567079544, + "weighted_rkd_loss": 0.00693986255209893 + }, + { + "epoch": 5.728643216080402, + "flow_matching_loss": 0.015781704988330602, + "grad_norm": 0.17194882035255432, + "learning_rate": 7.47996168276856e-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": 0.004093670891597867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015222638729028403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009709605574607849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015781704988330602, + "weighted_rkd_loss": 0.004854802787303924 + }, + { + "epoch": 5.745393634840871, + "flow_matching_loss": 0.01642342172563076, + "grad_norm": 0.14569294452667236, + "learning_rate": 7.394286625246258e-06, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00585379155818373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002785720257088542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014493303652852774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01642342172563076, + "weighted_rkd_loss": 0.007246651826426387 + }, + { + "epoch": 5.76214405360134, + "flow_matching_loss": 0.016525352746248244, + "grad_norm": 0.1307617723941803, + "learning_rate": 7.308944266266063e-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": 0.005043041077442467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002132825099397451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012218907242640853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016525352746248244, + "weighted_rkd_loss": 0.0061094536213204265 + }, + { + "epoch": 5.778894472361809, + "flow_matching_loss": 0.016573724709451197, + "grad_norm": 0.13970951735973358, + "learning_rate": 7.223938338979233e-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": 0.0046006364049389955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001974192395573482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011175465257838368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016573724709451197, + "weighted_rkd_loss": 0.005587732628919184 + }, + { + "epoch": 5.795644891122278, + "flow_matching_loss": 0.018552876077592373, + "grad_norm": 0.16639065742492676, + "learning_rate": 7.139272561820397e-06, + "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": 0.005241074599325657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020407545613124966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012522903643548489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018552876077592373, + "weighted_rkd_loss": 0.0062614518217742445 + }, + { + "epoch": 5.812395309882747, + "flow_matching_loss": 0.014944805391132831, + "grad_norm": 0.18595440685749054, + "learning_rate": 7.054950638344936e-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": 0.004425915516912937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00172965683741495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010581487929448486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014944805391132831, + "weighted_rkd_loss": 0.005290743964724243 + }, + { + "epoch": 5.8291457286432165, + "flow_matching_loss": 0.016345801949501037, + "grad_norm": 0.1472901552915573, + "learning_rate": 6.970976257066941e-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": 0.0051440172363072635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020479016355238857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012335935980081559, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016345801949501037, + "weighted_rkd_loss": 0.006167967990040779 + }, + { + "epoch": 5.8458961474036855, + "flow_matching_loss": 0.01786613268777728, + "grad_norm": 0.1330331712961197, + "learning_rate": 6.887353091297896e-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": 0.00458008588757366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001994237513281405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011154409311711789, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01786613268777728, + "weighted_rkd_loss": 0.0055772046558558944 + }, + { + "epoch": 5.8626465661641545, + "flow_matching_loss": 0.01581496661528945, + "grad_norm": 0.1423383206129074, + "learning_rate": 6.804084798985965e-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": 0.004506601137109101, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019407646264880895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010953966854140162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01581496661528945, + "weighted_rkd_loss": 0.005476983427070081 + }, + { + "epoch": 5.8793969849246235, + "flow_matching_loss": 0.017252940963953734, + "grad_norm": 0.20213234424591064, + "learning_rate": 6.721175022556009e-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": 0.004888100852258504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019750437000766396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01175124542787671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017252940963953734, + "weighted_rkd_loss": 0.005875622713938355 + }, + { + "epoch": 5.8961474036850925, + "flow_matching_loss": 0.016892566438764335, + "grad_norm": 0.17500615119934082, + "learning_rate": 6.638627388750245e-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": 0.005072918278165162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002193704899400473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012339541455730796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016892566438764335, + "weighted_rkd_loss": 0.006169770727865398 + }, + { + "epoch": 5.9128978224455615, + "flow_matching_loss": 0.018169266637414694, + "grad_norm": 0.1558530330657959, + "learning_rate": 6.556445508469583e-06, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004949927981942892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019603127497248352, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01186016877181828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018169266637414694, + "weighted_rkd_loss": 0.00593008438590914 + }, + { + "epoch": 5.9296482412060305, + "flow_matching_loss": 0.01648208061233163, + "grad_norm": 0.13735370337963104, + "learning_rate": 6.474632976615708e-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": 0.004612542688846588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002068738080561161, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01129382341168821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01648208061233163, + "weighted_rkd_loss": 0.005646911705844105 + }, + { + "epoch": 5.9463986599664995, + "flow_matching_loss": 0.016126606985926627, + "grad_norm": 0.15549711883068085, + "learning_rate": 6.393193371933791e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004885917692445218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001710111042484641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011481946427375078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016126606985926627, + "weighted_rkd_loss": 0.005740973213687539 + }, + { + "epoch": 5.9631490787269685, + "flow_matching_loss": 0.015177852846682072, + "grad_norm": 0.1654539257287979, + "learning_rate": 6.312130256855977e-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": 0.004428159282542765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001902568421792239, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010758886951953173, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015177852846682072, + "weighted_rkd_loss": 0.0053794434759765865 + }, + { + "epoch": 5.9798994974874375, + "flow_matching_loss": 0.015710619650781153, + "grad_norm": 0.1425623595714569, + "learning_rate": 6.23144717734552e-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": 0.003932594321668148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016364209004677832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009501609532162548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015710619650781153, + "weighted_rkd_loss": 0.004750804766081274 + }, + { + "epoch": 5.9966499162479066, + "flow_matching_loss": 0.017744600120931864, + "grad_norm": 0.20558246970176697, + "learning_rate": 6.151147662741695e-06, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051352317910641435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002128293039277196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01239875671453774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017744600120931864, + "weighted_rkd_loss": 0.00619937835726887 + }, + { + "epoch": 6.013400335008376, + "flow_matching_loss": 0.016692350897938014, + "grad_norm": 0.15531615912914276, + "learning_rate": 6.071235225605415e-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": 0.0044694194104522465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018079525209032, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010746791213750839, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016692350897938014, + "weighted_rkd_loss": 0.005373395606875419 + }, + { + "epoch": 6.030150753768845, + "flow_matching_loss": 0.015450216457247734, + "grad_norm": 0.17744119465351105, + "learning_rate": 5.991713361565546e-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": 0.004569414188154042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018782433704473077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011017071735113858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015450216457247734, + "weighted_rkd_loss": 0.005508535867556929 + }, + { + "epoch": 6.046901172529314, + "flow_matching_loss": 0.015183007437735796, + "grad_norm": 0.21500062942504883, + "learning_rate": 5.912585549166051e-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": 0.004672595020383597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019326795474626124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011277869530022144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015183007437735796, + "weighted_rkd_loss": 0.005638934765011072 + }, + { + "epoch": 6.063651591289783, + "flow_matching_loss": 0.01673786323517561, + "grad_norm": 0.15154196321964264, + "learning_rate": 5.833855249713777e-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": 0.004518775991164148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00174813725752756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010785689298063516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01673786323517561, + "weighted_rkd_loss": 0.005392844649031758 + }, + { + "epoch": 6.080402010050252, + "flow_matching_loss": 0.016367413848638535, + "grad_norm": 0.15717095136642456, + "learning_rate": 5.7555259071270645e-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": 0.004585496545769274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002205381088424474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01137637412175536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016367413848638535, + "weighted_rkd_loss": 0.00568818706087768 + }, + { + "epoch": 6.097152428810721, + "flow_matching_loss": 0.015005042776465416, + "grad_norm": 0.13135921955108643, + "learning_rate": 5.677600947785117e-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": 0.003971590660512447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001454745524097234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009397926880046726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015005042776465416, + "weighted_rkd_loss": 0.004698963440023363 + }, + { + "epoch": 6.11390284757119, + "flow_matching_loss": 0.01685640923678875, + "grad_norm": 0.18737980723381042, + "learning_rate": 5.600083780378099e-06, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005439688265323639, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002393451309762895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013272827863693238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01685640923678875, + "weighted_rkd_loss": 0.006636413931846619 + }, + { + "epoch": 6.130653266331659, + "flow_matching_loss": 0.016260798461735247, + "grad_norm": 0.13233205676078796, + "learning_rate": 5.522977795758038e-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": 0.005148716387338936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002030079159885645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012327512074261904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016260798461735247, + "weighted_rkd_loss": 0.006163756037130952 + }, + { + "epoch": 6.147403685092128, + "flow_matching_loss": 0.0165544499643147, + "grad_norm": 0.19198189675807953, + "learning_rate": 5.4462863667904805e-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": 0.004077224410139024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020296421251259746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01018409114331007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0165544499643147, + "weighted_rkd_loss": 0.005092045571655035 + }, + { + "epoch": 6.164154103852597, + "flow_matching_loss": 0.017229913640767335, + "grad_norm": 0.1838519275188446, + "learning_rate": 5.370012848206983e-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": 0.004221946839243174, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017063068808056414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010150200640782713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017229913640767335, + "weighted_rkd_loss": 0.005075100320391357 + }, + { + "epoch": 6.180904522613066, + "flow_matching_loss": 0.016627094615250824, + "grad_norm": 0.14189720153808594, + "learning_rate": 5.29416057645834e-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": 0.004839239851571619, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002113574917893857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011792054679244757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016627094615250824, + "weighted_rkd_loss": 0.005896027339622379 + }, + { + "epoch": 6.197654941373535, + "flow_matching_loss": 0.016476182825863363, + "grad_norm": 0.14053557813167572, + "learning_rate": 5.218732869568635e-06, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00519660187419504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023840540554374455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012777257803827524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016476182825863363, + "weighted_rkd_loss": 0.006388628901913762 + }, + { + "epoch": 6.214405360134004, + "flow_matching_loss": 0.015468492917716504, + "grad_norm": 0.1435490995645523, + "learning_rate": 5.14373302699013e-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": 0.004912928817793727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018873865308705717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01171324416063726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015468492917716504, + "weighted_rkd_loss": 0.00585662208031863 + }, + { + "epoch": 6.231155778894473, + "flow_matching_loss": 0.015904262941330672, + "grad_norm": 0.15239299833774567, + "learning_rate": 5.069164329458894e-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": 0.004692935664206743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018197374884039164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011205608956515789, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015904262941330672, + "weighted_rkd_loss": 0.0056028044782578945 + }, + { + "epoch": 6.247906197654942, + "flow_matching_loss": 0.017712344974279405, + "grad_norm": 0.17586107552051544, + "learning_rate": 4.995030038851342e-06, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004220141563564539, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018938871507998556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010334170144051313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017712344974279405, + "weighted_rkd_loss": 0.005167085072025656 + }, + { + "epoch": 6.264656616415411, + "flow_matching_loss": 0.015475019440054893, + "grad_norm": 0.18944132328033447, + "learning_rate": 4.9213333980415e-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": 0.004483212856575847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001734148757532239, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010700574517250061, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015475019440054893, + "weighted_rkd_loss": 0.005350287258625031 + }, + { + "epoch": 6.28140703517588, + "flow_matching_loss": 0.016900824569165706, + "grad_norm": 0.16883336007595062, + "learning_rate": 4.848077630759175e-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": 0.00466771477367729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019422173150815069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011277646850794553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016900824569165706, + "weighted_rkd_loss": 0.005638823425397277 + }, + { + "epoch": 6.298157453936349, + "flow_matching_loss": 0.015882111340761184, + "grad_norm": 0.1701260656118393, + "learning_rate": 4.775265941448951e-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": 0.004563337680883706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017544721951708198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010881147533655166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015882111340761184, + "weighted_rkd_loss": 0.005440573766827583 + }, + { + "epoch": 6.314907872696818, + "flow_matching_loss": 0.01612699143588543, + "grad_norm": 0.17910315096378326, + "learning_rate": 4.702901515129989e-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": 0.004781160946004092, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018259460222907365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011388267949223518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01612699143588543, + "weighted_rkd_loss": 0.005694133974611759 + }, + { + "epoch": 6.331658291457287, + "flow_matching_loss": 0.01585171017795801, + "grad_norm": 0.1363685131072998, + "learning_rate": 4.630987517256734e-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": 0.004472851287573576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016937122913077475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010639414750039577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01585171017795801, + "weighted_rkd_loss": 0.005319707375019789 + }, + { + "epoch": 6.348408710217756, + "flow_matching_loss": 0.015469970647245646, + "grad_norm": 0.13690084218978882, + "learning_rate": 4.559527093580412e-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": 0.004476161859929562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019692603673320264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010921584069728851, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015469970647245646, + "weighted_rkd_loss": 0.0054607920348644255 + }, + { + "epoch": 6.365159128978225, + "flow_matching_loss": 0.017418408021330833, + "grad_norm": 0.1333802491426468, + "learning_rate": 4.488523370011456e-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": 0.004803127376362681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020921259187161923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011698380671441555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017418408021330833, + "weighted_rkd_loss": 0.005849190335720777 + }, + { + "epoch": 6.381909547738694, + "flow_matching_loss": 0.015745705273002387, + "grad_norm": 0.2102486789226532, + "learning_rate": 4.417979452482766e-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": 0.005005351454019547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002100074465852231, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012110777292400598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015745705273002387, + "weighted_rkd_loss": 0.006055388646200299 + }, + { + "epoch": 6.398659966499163, + "flow_matching_loss": 0.014893369190394879, + "grad_norm": 0.13255921006202698, + "learning_rate": 4.347898426813811e-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": 0.005195147078484297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002191038301680237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012581332447007298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014893369190394879, + "weighted_rkd_loss": 0.006290666223503649 + }, + { + "epoch": 6.415410385259632, + "flow_matching_loss": 0.015815038699656725, + "grad_norm": 0.1442941427230835, + "learning_rate": 4.278283358575694e-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": 0.004424917628057301, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00187108451500535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010720919771119952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015815038699656725, + "weighted_rkd_loss": 0.005360459885559976 + }, + { + "epoch": 6.432160804020101, + "flow_matching_loss": 0.016232730634510518, + "grad_norm": 0.1966170370578766, + "learning_rate": 4.2091372929570145e-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": 0.0037545856786891816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017783464048989118, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00928751789033413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016232730634510518, + "weighted_rkd_loss": 0.004643758945167065 + }, + { + "epoch": 6.44891122278057, + "flow_matching_loss": 0.0173671611584723, + "grad_norm": 0.16270123422145844, + "learning_rate": 4.1404632546306934e-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": 0.00426407353952527, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017715994967147709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010299746599048375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0173671611584723, + "weighted_rkd_loss": 0.005149873299524188 + }, + { + "epoch": 6.465661641541039, + "flow_matching_loss": 0.015224199555814267, + "grad_norm": 0.1519557237625122, + "learning_rate": 4.0722642476216286e-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": 0.004490367136895657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019021688785869628, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010882903169840575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015224199555814267, + "weighted_rkd_loss": 0.0054414515849202875 + }, + { + "epoch": 6.482412060301508, + "flow_matching_loss": 0.01661739433184266, + "grad_norm": 0.14995500445365906, + "learning_rate": 4.00454325517533e-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": 0.004987293970771134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022416375228203832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012216225545853377, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01661739433184266, + "weighted_rkd_loss": 0.0061081127729266885 + }, + { + "epoch": 6.499162479061977, + "flow_matching_loss": 0.015825352538377045, + "grad_norm": 0.16931673884391785, + "learning_rate": 3.937303239627393e-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": 0.004396289819851518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001814883027691394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010607462516054511, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015825352538377045, + "weighted_rkd_loss": 0.005303731258027256 + }, + { + "epoch": 6.515912897822446, + "flow_matching_loss": 0.01738193165510893, + "grad_norm": 0.14341159164905548, + "learning_rate": 3.870547142273918e-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": 0.0036826774710789325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013857228332199156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008751077763736247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01738193165510893, + "weighted_rkd_loss": 0.004375538881868124 + }, + { + "epoch": 6.532663316582915, + "flow_matching_loss": 0.01572351260110736, + "grad_norm": 0.16346746683120728, + "learning_rate": 3.8042778832428736e-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": 0.004531700373627245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017317167716100813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010795117449015379, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01572351260110736, + "weighted_rkd_loss": 0.005397558724507689 + }, + { + "epoch": 6.549413735343384, + "flow_matching_loss": 0.019746193196624517, + "grad_norm": 0.14955230057239532, + "learning_rate": 3.7384983613663195e-06, + "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": 0.004160749423317611, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018678586231544614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010189357539638877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.019746193196624517, + "weighted_rkd_loss": 0.005094678769819438 + }, + { + "epoch": 6.566164154103853, + "flow_matching_loss": 0.018092065956443547, + "grad_norm": 0.17253366112709045, + "learning_rate": 3.6732114540536503e-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": 0.004915033443830907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020113187143579125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011841385671868921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018092065956443547, + "weighted_rkd_loss": 0.0059206928359344605 + }, + { + "epoch": 6.582914572864322, + "flow_matching_loss": 0.017865691613405944, + "grad_norm": 0.21846531331539154, + "learning_rate": 3.608420017165683e-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": 0.004521297919563949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002094288810621947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011136884754523634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017865691613405944, + "weighted_rkd_loss": 0.005568442377261817 + }, + { + "epoch": 6.599664991624791, + "flow_matching_loss": 0.016273113992065193, + "grad_norm": 0.14731267094612122, + "learning_rate": 3.5441268848897664e-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": 0.004396083881147206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017308910959400237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010523058846592903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016273113992065193, + "weighted_rkd_loss": 0.005261529423296452 + }, + { + "epoch": 6.61641541038526, + "flow_matching_loss": 0.016229044739156963, + "grad_norm": 0.13433827459812164, + "learning_rate": 3.480334869615792e-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": 0.003977289563044906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018249677086714655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009779546782374382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016229044739156963, + "weighted_rkd_loss": 0.004889773391187191 + }, + { + "epoch": 6.633165829145729, + "flow_matching_loss": 0.01615925971418619, + "grad_norm": 0.17934858798980713, + "learning_rate": 3.417046761813155e-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": 0.004120925953611731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002054018119815737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010295870155096054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01615925971418619, + "weighted_rkd_loss": 0.005147935077548027 + }, + { + "epoch": 6.649916247906198, + "flow_matching_loss": 0.01634274534881115, + "grad_norm": 0.15029595792293549, + "learning_rate": 3.3542653299087263e-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": 0.0038260640110820534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001958185143303126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009610313083976507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01634274534881115, + "weighted_rkd_loss": 0.004805156541988253 + }, + { + "epoch": 6.666666666666667, + "flow_matching_loss": 0.01320421127602458, + "grad_norm": 0.16808393597602844, + "learning_rate": 3.2919933201657117e-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": 0.004408310819417238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018108319956809283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010627453541383147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01320421127602458, + "weighted_rkd_loss": 0.0053137267706915734 + }, + { + "epoch": 6.683417085427136, + "flow_matching_loss": 0.01695033786818385, + "grad_norm": 0.175664022564888, + "learning_rate": 3.2302334565635584e-06, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004060668661259115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016366543364711105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009757991693913937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01695033786818385, + "weighted_rkd_loss": 0.004878995846956969 + }, + { + "epoch": 6.700167504187605, + "flow_matching_loss": 0.014768802467733622, + "grad_norm": 0.19610930979251862, + "learning_rate": 3.1689884406787674e-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": 0.0039969151839613914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018481418024748563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00984197212383151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014768802467733622, + "weighted_rkd_loss": 0.004920986061915755 + }, + { + "epoch": 6.716917922948074, + "flow_matching_loss": 0.01720321225002408, + "grad_norm": 0.16937272250652313, + "learning_rate": 3.1082609515667424e-06, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004821825376711786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019550187862478197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011598669551312923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01720321225002408, + "weighted_rkd_loss": 0.0057993347756564615 + }, + { + "epoch": 6.733668341708543, + "flow_matching_loss": 0.015799351502209902, + "grad_norm": 0.183858260512352, + "learning_rate": 3.048053645644594e-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": 0.004556529875844717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018371412064880134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010950201004743577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015799351502209902, + "weighted_rkd_loss": 0.005475100502371788 + }, + { + "epoch": 6.750418760469012, + "flow_matching_loss": 0.01648924229666591, + "grad_norm": 0.1715034395456314, + "learning_rate": 2.988369156574916e-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": 0.004184935172088444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016394643927924335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010009334795176983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01648924229666591, + "weighted_rkd_loss": 0.005004667397588492 + }, + { + "epoch": 6.767169179229481, + "flow_matching_loss": 0.016042891331017016, + "grad_norm": 0.13695967197418213, + "learning_rate": 2.9292100951506225e-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": 0.004686999879777431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018519218778237701, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01122592156752944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016042891331017016, + "weighted_rkd_loss": 0.00561296078376472 + }, + { + "epoch": 6.78391959798995, + "flow_matching_loss": 0.016733720898628235, + "grad_norm": 0.16754966974258423, + "learning_rate": 2.8705790491807033e-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": 0.0039164737798273565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017174503766000271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009550398029386998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016733720898628235, + "weighted_rkd_loss": 0.004775199014693499 + }, + { + "epoch": 6.800670016750419, + "flow_matching_loss": 0.014544985536485911, + "grad_norm": 0.15087449550628662, + "learning_rate": 2.8124785833770433e-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": 0.00489426520653069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020002575591206552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011788788065314294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014544985536485911, + "weighted_rkd_loss": 0.005894394032657147 + }, + { + "epoch": 6.817420435510888, + "flow_matching_loss": 0.016845161374658347, + "grad_norm": 0.18539758026599884, + "learning_rate": 2.754911239242241e-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": 0.004746794002130628, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021222085459157826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011615796573460102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016845161374658347, + "weighted_rkd_loss": 0.005807898286730051 + }, + { + "epoch": 6.834170854271357, + "flow_matching_loss": 0.01456131674349308, + "grad_norm": 0.17953617870807648, + "learning_rate": 2.6978795349584204e-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": 0.00482990201562643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018176681362092495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011477472074329854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01456131674349308, + "weighted_rkd_loss": 0.005738736037164927 + }, + { + "epoch": 6.850921273031826, + "flow_matching_loss": 0.016708610765635967, + "grad_norm": 0.1707935482263565, + "learning_rate": 2.641385965277088e-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": 0.004654878587462008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021587817347608505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0114685389213264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016708610765635967, + "weighted_rkd_loss": 0.0057342694606632 + }, + { + "epoch": 6.867671691792295, + "flow_matching_loss": 0.014694995246827602, + "grad_norm": 0.19747313857078552, + "learning_rate": 2.5854330014099923e-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": 0.0036429993459023537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015449580852873624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008830956788733602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014694995246827602, + "weighted_rkd_loss": 0.004415478394366801 + }, + { + "epoch": 6.884422110552764, + "flow_matching_loss": 0.01677362499758601, + "grad_norm": 0.12381528317928314, + "learning_rate": 2.5300230909210298e-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": 0.003627436398528516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018569727661088109, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009111845586448907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01677362499758601, + "weighted_rkd_loss": 0.004555922793224454 + }, + { + "epoch": 6.901172529313233, + "flow_matching_loss": 0.015464589465409517, + "grad_norm": 0.1643161177635193, + "learning_rate": 2.4751586576191983e-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": 0.00446525604929775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019164797675330191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010846991883590817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015464589465409517, + "weighted_rkd_loss": 0.0054234959417954086 + }, + { + "epoch": 6.917922948073702, + "flow_matching_loss": 0.015309206023812295, + "grad_norm": 0.1433928906917572, + "learning_rate": 2.420842101452538e-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": 0.004670693026855588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016811168286949396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011022503068670631, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015309206023812295, + "weighted_rkd_loss": 0.005511251534335316 + }, + { + "epoch": 6.934673366834171, + "flow_matching_loss": 0.015999872423708438, + "grad_norm": 0.158779114484787, + "learning_rate": 2.3670757984031816e-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": 0.004784934082999825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001908725016983226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011478593293577432, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015999872423708438, + "weighted_rkd_loss": 0.005739296646788716 + }, + { + "epoch": 6.95142378559464, + "flow_matching_loss": 0.0146644729655236, + "grad_norm": 0.14927701652050018, + "learning_rate": 2.313862100383404e-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": 0.003657519840635359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001735816220752895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009050855925306679, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0146644729655236, + "weighted_rkd_loss": 0.004525427962653339 + }, + { + "epoch": 6.968174204355109, + "flow_matching_loss": 0.01477276049554348, + "grad_norm": 0.18146266043186188, + "learning_rate": 2.2612033351327406e-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": 0.00446374686434865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017345455940812825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010662039229646324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01477276049554348, + "weighted_rkd_loss": 0.005331019614823162 + }, + { + "epoch": 6.984924623115578, + "flow_matching_loss": 0.01612620856612921, + "grad_norm": 0.1247067078948021, + "learning_rate": 2.2091018061161788e-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": 0.004328918852843344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019952588831074537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01065309657715261, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01612620856612921, + "weighted_rkd_loss": 0.005326548288576305 + }, + { + "epoch": 7.001675041876047, + "flow_matching_loss": 0.01582100810483098, + "grad_norm": 0.1982302963733673, + "learning_rate": 2.15755979242338e-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": 0.004648764641024173, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021950488444417715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011492578079923987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01582100810483098, + "weighted_rkd_loss": 0.005746289039961994 + }, + { + "epoch": 7.018425460636516, + "flow_matching_loss": 0.018236480932682753, + "grad_norm": 0.19694870710372925, + "learning_rate": 2.106579548669e-06, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005001381016336381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020459327613934876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012048694863915443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018236480932682753, + "weighted_rkd_loss": 0.0060243474319577215 + }, + { + "epoch": 7.035175879396985, + "flow_matching_loss": 0.014276114106178284, + "grad_norm": 0.1846606433391571, + "learning_rate": 2.0561633048940477e-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": 0.004038699180819094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016007982019800692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009678196534514428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014276114106178284, + "weighted_rkd_loss": 0.004839098267257214 + }, + { + "epoch": 7.051926298157454, + "flow_matching_loss": 0.016784136835485696, + "grad_norm": 0.15324676036834717, + "learning_rate": 2.006313266468355e-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": 0.004523446527309715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001994902826845646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011041795928031206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016784136835485696, + "weighted_rkd_loss": 0.005520897964015603 + }, + { + "epoch": 7.068676716917923, + "flow_matching_loss": 0.014370397850871085, + "grad_norm": 0.16374434530735016, + "learning_rate": 1.957031613994097e-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": 0.0038065645145252346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014738554309587926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009086984395980834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014370397850871085, + "weighted_rkd_loss": 0.004543492197990417 + }, + { + "epoch": 7.085427135678392, + "flow_matching_loss": 0.01604610849171877, + "grad_norm": 0.13959617912769318, + "learning_rate": 1.9083205032103995e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0047938420437276365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002127973339520395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011715657310560346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01604610849171877, + "weighted_rkd_loss": 0.005857828655280173 + }, + { + "epoch": 7.102177554438861, + "flow_matching_loss": 0.01643010778352618, + "grad_norm": 0.15431803464889526, + "learning_rate": 1.8601820648990536e-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": 0.003699658077675849, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016984055924694984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009097721613943576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01643010778352618, + "weighted_rkd_loss": 0.004548860806971788 + }, + { + "epoch": 7.11892797319933, + "flow_matching_loss": 0.014956025872379541, + "grad_norm": 0.15228129923343658, + "learning_rate": 1.8126184047912936e-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": 0.004501348035410047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00216407316038385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011166769126430155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014956025872379541, + "weighted_rkd_loss": 0.005583384563215077 + }, + { + "epoch": 7.135678391959799, + "flow_matching_loss": 0.015140857826918364, + "grad_norm": 0.1639655977487564, + "learning_rate": 1.7656316034757024e-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": 0.004590366035699844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018575215362943709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011038253549486399, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015140857826918364, + "weighted_rkd_loss": 0.005519126774743199 + }, + { + "epoch": 7.152428810720268, + "flow_matching_loss": 0.014944946672767401, + "grad_norm": 0.14010633528232574, + "learning_rate": 1.7192237163071816e-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": 0.004896424198523164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002066554082557559, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011859402479603886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014944946672767401, + "weighted_rkd_loss": 0.005929701239801943 + }, + { + "epoch": 7.169179229480737, + "flow_matching_loss": 0.016208091285079718, + "grad_norm": 0.16711357235908508, + "learning_rate": 1.6733967733170573e-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": 0.0037203690968453883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015936146199237555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009034352796152233, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016208091285079718, + "weighted_rkd_loss": 0.004517176398076117 + }, + { + "epoch": 7.185929648241206, + "flow_matching_loss": 0.014888655114918947, + "grad_norm": 0.15203069150447845, + "learning_rate": 1.628152779124279e-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": 0.004339498141780496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001999542448902503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010678538680076599, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014888655114918947, + "weighted_rkd_loss": 0.005339269340038299 + }, + { + "epoch": 7.202680067001675, + "flow_matching_loss": 0.014624554850161075, + "grad_norm": 0.11805262416601181, + "learning_rate": 1.583493712847715e-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": 0.0042874168837443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001616271329112351, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010191105119884015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014624554850161075, + "weighted_rkd_loss": 0.005095552559942007 + }, + { + "epoch": 7.219430485762144, + "flow_matching_loss": 0.017096712067723275, + "grad_norm": 0.13011588156223297, + "learning_rate": 1.5394215280196084e-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": 0.004473495739512145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017709834850393237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01071797488257289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017096712067723275, + "weighted_rkd_loss": 0.005358987441286445 + }, + { + "epoch": 7.236180904522613, + "flow_matching_loss": 0.015585697442293166, + "grad_norm": 0.18555836379528046, + "learning_rate": 1.4959381525000904e-06, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004154981835745275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016431983793154358, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009953161980956794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015585697442293166, + "weighted_rkd_loss": 0.004976580990478397 + }, + { + "epoch": 7.252931323283082, + "flow_matching_loss": 0.016454843897372486, + "grad_norm": 0.14526554942131042, + "learning_rate": 1.453045488392879e-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": 0.004315899196080863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001923052000347525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010554850520566105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016454843897372486, + "weighted_rkd_loss": 0.005277425260283053 + }, + { + "epoch": 7.269681742043551, + "flow_matching_loss": 0.015291227679699659, + "grad_norm": 0.12427052855491638, + "learning_rate": 1.4107454119620484e-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": 0.0038240417605265977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014376269245985895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00908571039326489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015291227679699659, + "weighted_rkd_loss": 0.004542855196632445 + }, + { + "epoch": 7.28643216080402, + "flow_matching_loss": 0.015831883577629925, + "grad_norm": 0.1394655555486679, + "learning_rate": 1.3690397735499778e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004891686118207872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002289877535076812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012073249882087111, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015831883577629925, + "weighted_rkd_loss": 0.0060366249410435556 + }, + { + "epoch": 7.303182579564489, + "flow_matching_loss": 0.01743771005421877, + "grad_norm": 0.14378996193408966, + "learning_rate": 1.327930397496403e-06, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005148415197618305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002286371891386807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012583202309906482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01743771005421877, + "weighted_rkd_loss": 0.006291601154953241 + }, + { + "epoch": 7.319932998324958, + "flow_matching_loss": 0.013519060239195824, + "grad_norm": 0.2160998433828354, + "learning_rate": 1.287419082058604e-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": 0.004035786679014564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019525676732882857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01002414091490209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013519060239195824, + "weighted_rkd_loss": 0.005012070457451045 + }, + { + "epoch": 7.336683417085427, + "flow_matching_loss": 0.015327712800353766, + "grad_norm": 0.1508062183856964, + "learning_rate": 1.2475075993327589e-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": 0.0042715171119198205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016422059037722648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010185240115970374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015327712800353766, + "weighted_rkd_loss": 0.005092620057985187 + }, + { + "epoch": 7.353433835845896, + "flow_matching_loss": 0.015343380253762006, + "grad_norm": 0.15916304290294647, + "learning_rate": 1.208197695176413e-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": 0.0046300238231197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002693808858748525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011953856470063329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015343380253762006, + "weighted_rkd_loss": 0.0059769282350316645 + }, + { + "epoch": 7.370184254606365, + "flow_matching_loss": 0.01437976472079754, + "grad_norm": 0.14108532667160034, + "learning_rate": 1.1694910891321259e-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": 0.004367586877197027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019202484982088208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010655422182753681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01437976472079754, + "weighted_rkd_loss": 0.005327711091376841 + }, + { + "epoch": 7.386934673366834, + "flow_matching_loss": 0.015470136143267154, + "grad_norm": 0.15935403108596802, + "learning_rate": 1.1313894743522363e-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": 0.003939059865660966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001673631521407515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009551751287654042, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015470136143267154, + "weighted_rkd_loss": 0.004775875643827021 + }, + { + "epoch": 7.403685092127303, + "flow_matching_loss": 0.016630494873970746, + "grad_norm": 0.20485074818134308, + "learning_rate": 1.0938945175248016e-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": 0.004128277907148004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001766240206779912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01002279594540596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016630494873970746, + "weighted_rkd_loss": 0.00501139797270298 + }, + { + "epoch": 7.420435510887772, + "flow_matching_loss": 0.01638146713376045, + "grad_norm": 0.10602868348360062, + "learning_rate": 1.0570078588007086e-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": 0.004238395742140711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017436613445170224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010220452863723039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01638146713376045, + "weighted_rkd_loss": 0.0051102264318615195 + }, + { + "epoch": 7.437185929648241, + "flow_matching_loss": 0.014488376211374998, + "grad_norm": 0.12349338829517365, + "learning_rate": 1.0207311117219009e-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": 0.0041768827708438035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017094228183850645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01006318824365735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014488376211374998, + "weighted_rkd_loss": 0.005031594121828675 + }, + { + "epoch": 7.45393634840871, + "flow_matching_loss": 0.014135436434298754, + "grad_norm": 0.11960218101739883, + "learning_rate": 9.850658631508198e-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": 0.003858961584046483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016201299498789013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009338053129613399, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014135436434298754, + "weighted_rkd_loss": 0.004669026564806699 + }, + { + "epoch": 7.4706867671691795, + "flow_matching_loss": 0.015964719746261836, + "grad_norm": 0.15216432511806488, + "learning_rate": 9.500136732009751e-07, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038564404007047416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00145489246933721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009167773276567459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015964719746261836, + "weighted_rkd_loss": 0.004583886638283729 + }, + { + "epoch": 7.4874371859296485, + "flow_matching_loss": 0.013642612006515264, + "grad_norm": 0.14385931193828583, + "learning_rate": 9.155760751687054e-07, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039043609518557786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017102246696595103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009518946800380946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013642612006515264, + "weighted_rkd_loss": 0.004759473400190473 + }, + { + "epoch": 7.5041876046901175, + "flow_matching_loss": 0.014753709733486175, + "grad_norm": 0.16965070366859436, + "learning_rate": 8.817545754661155e-07, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00408722497522831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001713181775994599, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009887631935998797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014753709733486175, + "weighted_rkd_loss": 0.004943815967999398 + }, + { + "epoch": 7.5209380234505865, + "flow_matching_loss": 0.016744329128414392, + "grad_norm": 0.16780605912208557, + "learning_rate": 8.485506535551668e-07, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004479729966260493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019441045355051757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01090356446802616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016744329128414392, + "weighted_rkd_loss": 0.00545178223401308 + }, + { + "epoch": 7.5376884422110555, + "flow_matching_loss": 0.014030791632831097, + "grad_norm": 0.1584310531616211, + "learning_rate": 8.159657618829685e-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": 0.003778664767742157, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016007275553420186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009158057114109396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014030791632831097, + "weighted_rkd_loss": 0.004579028557054698 + }, + { + "epoch": 7.5544388609715245, + "flow_matching_loss": 0.016403439175337552, + "grad_norm": 0.14649353921413422, + "learning_rate": 7.840013258182471e-07, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003972286405041814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014566533092875034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009401226090267301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016403439175337552, + "weighted_rkd_loss": 0.0047006130451336505 + }, + { + "epoch": 7.5711892797319935, + "flow_matching_loss": 0.016914738900959493, + "grad_norm": 0.12157639861106873, + "learning_rate": 7.526587435889809e-07, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003476941166445613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013800669636111706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008333949279040099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016914738900959493, + "weighted_rkd_loss": 0.004166974639520049 + }, + { + "epoch": 7.5879396984924625, + "flow_matching_loss": 0.01702210819348693, + "grad_norm": 0.16324695944786072, + "learning_rate": 7.219393862212593e-07, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004177236021496356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001766754686832428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010121226636692881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01702210819348693, + "weighted_rkd_loss": 0.005060613318346441 + }, + { + "epoch": 7.6046901172529315, + "flow_matching_loss": 0.016736098565161228, + "grad_norm": 0.13862305879592896, + "learning_rate": 6.918445974792819e-07, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035174431046471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017616379889659583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00879652420990169, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016736098565161228, + "weighted_rkd_loss": 0.004398262104950845 + }, + { + "epoch": 7.6214405360134005, + "flow_matching_loss": 0.015986019000411033, + "grad_norm": 0.1334938406944275, + "learning_rate": 6.6237569380661e-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": 0.003616045939270407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001374229503562674, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008606321271508932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015986019000411033, + "weighted_rkd_loss": 0.004303160635754466 + }, + { + "epoch": 7.63819095477387, + "flow_matching_loss": 0.01590715404599905, + "grad_norm": 0.14786288142204285, + "learning_rate": 6.335339642685495e-07, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004220075206831098, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017576340294908732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01019778442569077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01590715404599905, + "weighted_rkd_loss": 0.005098892212845385 + }, + { + "epoch": 7.654941373534339, + "flow_matching_loss": 0.015476820152252912, + "grad_norm": 0.1825518161058426, + "learning_rate": 6.05320670495787e-07, + "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": 0.0040603889618068935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017880713276099414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009908849187195302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015476820152252912, + "weighted_rkd_loss": 0.004954424593597651 + }, + { + "epoch": 7.671691792294808, + "flow_matching_loss": 0.014729360304772853, + "grad_norm": 0.18547452986240387, + "learning_rate": 5.777370466291954e-07, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004903944488614797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002282101463060826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012089990358799697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014729360304772853, + "weighted_rkd_loss": 0.006044995179399848 + }, + { + "epoch": 7.688442211055277, + "flow_matching_loss": 0.01499204495921731, + "grad_norm": 0.17845290899276733, + "learning_rate": 5.507842992658336e-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": 0.003743468993343413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015168702288065107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00900380820967257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01499204495921731, + "weighted_rkd_loss": 0.004501904104836285 + }, + { + "epoch": 7.705192629815746, + "flow_matching_loss": 0.01654659826308489, + "grad_norm": 0.19419962167739868, + "learning_rate": 5.24463607406192e-07, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004446967015974223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017932968446984887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010687231039628386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01654659826308489, + "weighted_rkd_loss": 0.005343615519814193 + }, + { + "epoch": 7.721943048576215, + "flow_matching_loss": 0.014629477728158235, + "grad_norm": 0.16444142162799835, + "learning_rate": 4.987761224025944e-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": 0.004281575768254697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017087026499211788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010271854186430573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014629477728158235, + "weighted_rkd_loss": 0.005135927093215286 + }, + { + "epoch": 7.738693467336684, + "flow_matching_loss": 0.01777318762615323, + "grad_norm": 0.13727985322475433, + "learning_rate": 4.737229679088556e-07, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036333360243588687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016103479429148138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008877020049840211, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01777318762615323, + "weighted_rkd_loss": 0.004438510024920106 + }, + { + "epoch": 7.755443886097153, + "flow_matching_loss": 0.016660143435001374, + "grad_norm": 0.13840682804584503, + "learning_rate": 4.493052398311115e-07, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004052148922346533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018098844622727484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009914182359352708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016660143435001374, + "weighted_rkd_loss": 0.004957091179676354 + }, + { + "epoch": 7.772194304857622, + "flow_matching_loss": 0.015904311370104552, + "grad_norm": 0.18793518841266632, + "learning_rate": 4.2552400627989043e-07, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004106870060786605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017587932874448598, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009972533490508795, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015904311370104552, + "weighted_rkd_loss": 0.004986266745254398 + }, + { + "epoch": 7.788944723618091, + "flow_matching_loss": 0.015392695367336274, + "grad_norm": 0.22930394113063812, + "learning_rate": 4.023803075233923e-07, + "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": 0.004883068730123341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002066736004780978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011832873662933707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015392695367336274, + "weighted_rkd_loss": 0.0059164368314668534 + }, + { + "epoch": 7.80569514237856, + "flow_matching_loss": 0.01676181796938181, + "grad_norm": 0.1597457081079483, + "learning_rate": 3.7987515594196997e-07, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036124582402408123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001387284556403756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00861220108345151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01676181796938181, + "weighted_rkd_loss": 0.004306100541725755 + }, + { + "epoch": 7.822445561139029, + "flow_matching_loss": 0.015276381466537713, + "grad_norm": 0.22032739222049713, + "learning_rate": 3.580095359838648e-07, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004534202092327177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016930141719058156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010761418426409364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015276381466537713, + "weighted_rkd_loss": 0.005380709213204682 + }, + { + "epoch": 7.839195979899498, + "flow_matching_loss": 0.015883130393922328, + "grad_norm": 0.19140352308750153, + "learning_rate": 3.3678440412212576e-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": 0.003926279791630805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016084841859992593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009461043868213892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015883130393922328, + "weighted_rkd_loss": 0.004730521934106946 + }, + { + "epoch": 7.855946398659967, + "flow_matching_loss": 0.016306853387504816, + "grad_norm": 0.12501484155654907, + "learning_rate": 3.162006888127833e-07, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004368531377986073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019060501595959066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010643112845718861, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016306853387504816, + "weighted_rkd_loss": 0.0053215564228594305 + }, + { + "epoch": 7.872696817420436, + "flow_matching_loss": 0.017385248746722936, + "grad_norm": 0.1701114922761917, + "learning_rate": 2.9625929045422327e-07, + "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": 0.00435904401820153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001872993551660329, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010591081576421857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017385248746722936, + "weighted_rkd_loss": 0.0052955407882109284 + }, + { + "epoch": 7.889447236180905, + "flow_matching_loss": 0.017262033838778734, + "grad_norm": 0.1395217478275299, + "learning_rate": 2.7696108134781186e-07, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036485470365732907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013480199966579675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008645114209502935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017262033838778734, + "weighted_rkd_loss": 0.004322557104751467 + }, + { + "epoch": 7.906197654941374, + "flow_matching_loss": 0.0179462818428874, + "grad_norm": 0.1584036499261856, + "learning_rate": 2.583069056597376e-07, + "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": 0.004363359813578427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018044174241367728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010531136952340602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0179462818428874, + "weighted_rkd_loss": 0.005265568476170301 + }, + { + "epoch": 7.922948073701843, + "flow_matching_loss": 0.014670177130028605, + "grad_norm": 0.16058886051177979, + "learning_rate": 2.4029757938407116e-07, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004476101510226727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019054161792155355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010857619298622011, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014670177130028605, + "weighted_rkd_loss": 0.005428809649311006 + }, + { + "epoch": 7.939698492462312, + "flow_matching_loss": 0.017554907780140638, + "grad_norm": 0.12912341952323914, + "learning_rate": 2.229338903070871e-07, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00388623287435621, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001478163164574653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009250628901645542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017554907780140638, + "weighted_rkd_loss": 0.004625314450822771 + }, + { + "epoch": 7.956448911222781, + "flow_matching_loss": 0.016205093823373318, + "grad_norm": 0.12646424770355225, + "learning_rate": 2.0621659797279645e-07, + "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": 0.0042540517169982195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017718847666401416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01027998821809888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016205093823373318, + "weighted_rkd_loss": 0.00513999410904944 + }, + { + "epoch": 7.97319932998325, + "flow_matching_loss": 0.016403627768158912, + "grad_norm": 0.18282198905944824, + "learning_rate": 1.901464336497233e-07, + "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": 0.004610209143720567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019326368696056307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011153055075556039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016403627768158912, + "weighted_rkd_loss": 0.0055765275377780196 + }, + { + "epoch": 7.989949748743719, + "flow_matching_loss": 0.015249314811080694, + "grad_norm": 0.17364180088043213, + "learning_rate": 1.7472410029891527e-07, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00389033502433449, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001546752726426348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009327422780916094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015249314811080694, + "weighted_rkd_loss": 0.004663711390458047 + }, + { + "epoch": 8.006700167504187, + "flow_matching_loss": 0.014920513145625591, + "grad_norm": 0.19061186909675598, + "learning_rate": 1.5995027254319173e-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": 0.0035726051311939953, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014776690921280533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008622879348695277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014920513145625591, + "weighted_rkd_loss": 0.004311439674347639 + }, + { + "epoch": 8.023450586264657, + "flow_matching_loss": 0.016243462171405553, + "grad_norm": 0.12961623072624207, + "learning_rate": 1.4582559663764372e-07, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038010131334885954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001989023806527257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00959105002693832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016243462171405553, + "weighted_rkd_loss": 0.00479552501346916 + }, + { + "epoch": 8.040201005025125, + "flow_matching_loss": 0.017470394633710383, + "grad_norm": 0.17603987455368042, + "learning_rate": 1.323506904413485e-07, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038247731747105716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016475262236781418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009297072421759367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017470394633710383, + "weighted_rkd_loss": 0.0046485362108796835 + }, + { + "epoch": 8.056951423785595, + "flow_matching_loss": 0.0145564878359437, + "grad_norm": 0.197255939245224, + "learning_rate": 1.195261433903594e-07, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042893028585240245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017500674701295793, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010328673152253031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.0145564878359437, + "weighted_rkd_loss": 0.005164336576126516 + }, + { + "epoch": 8.073701842546063, + "flow_matching_loss": 0.014916270691901445, + "grad_norm": 0.1560496985912323, + "learning_rate": 1.0735251647190969e-07, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003976433514617383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015865480876527728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009539415081962942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014916270691901445, + "weighted_rkd_loss": 0.004769707540981471 + }, + { + "epoch": 8.090452261306533, + "flow_matching_loss": 0.014834001008421183, + "grad_norm": 0.1427733153104782, + "learning_rate": 9.583034219987408e-08, + "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": 0.003930709883570671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015655665658414364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009426986379548908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014834001008421183, + "weighted_rkd_loss": 0.004713493189774454 + }, + { + "epoch": 8.107202680067001, + "flow_matching_loss": 0.013917479477822781, + "grad_norm": 0.1223326325416565, + "learning_rate": 8.496012459148717e-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": 0.0044130357215181, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002009857736993581, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010835929354652763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013917479477822781, + "weighted_rkd_loss": 0.0054179646773263816 + }, + { + "epoch": 8.12395309882747, + "flow_matching_loss": 0.017225508112460375, + "grad_norm": 0.11976644396781921, + "learning_rate": 7.474233914527618e-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": 0.003419135557487607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013502314046490937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00818850235082209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017225508112460375, + "weighted_rkd_loss": 0.004094251175411045 + }, + { + "epoch": 8.140703517587939, + "flow_matching_loss": 0.018462932761758567, + "grad_norm": 0.11892782896757126, + "learning_rate": 6.517743282028254e-08, + "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": 0.004395476914942265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017995877482462675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010590541595593094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.018462932761758567, + "weighted_rkd_loss": 0.005295270797796547 + }, + { + "epoch": 8.157453936348409, + "flow_matching_loss": 0.01574688982218504, + "grad_norm": 0.13970652222633362, + "learning_rate": 5.6265824016492604e-08, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004414934199303389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001722412754315883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010552281187847257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01574688982218504, + "weighted_rkd_loss": 0.0052761405939236285 + }, + { + "epoch": 8.174204355108877, + "flow_matching_loss": 0.01800237987190485, + "grad_norm": 0.12215915322303772, + "learning_rate": 4.800790255654719e-08, + "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": 0.003544078720733523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001673859596485272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00876201707869768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01800237987190485, + "weighted_rkd_loss": 0.00438100853934884 + }, + { + "epoch": 8.190954773869347, + "flow_matching_loss": 0.015847709495574236, + "grad_norm": 0.15015308558940887, + "learning_rate": 4.040402966867862e-08, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004173338948749006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015449339873157442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009891611896455288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015847709495574236, + "weighted_rkd_loss": 0.004945805948227644 + }, + { + "epoch": 8.207705192629815, + "flow_matching_loss": 0.015503469109535217, + "grad_norm": 0.20167481899261475, + "learning_rate": 3.3454537970925014e-08, + "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": 0.004341295640915632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021772518870420755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010859843250364066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015503469109535217, + "weighted_rkd_loss": 0.005429921625182033 + }, + { + "epoch": 8.224455611390285, + "flow_matching_loss": 0.01842154674232006, + "grad_norm": 0.19743748009204865, + "learning_rate": 2.715973145656192e-08, + "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": 0.0035309243947267534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015276472142431885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0085894959513098, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01842154674232006, + "weighted_rkd_loss": 0.0042947479756549 + }, + { + "epoch": 8.241206030150753, + "flow_matching_loss": 0.014870964922010899, + "grad_norm": 0.11341430246829987, + "learning_rate": 2.151988548082129e-08, + "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": 0.003632827172987163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014750438393093645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008740698220208287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014870964922010899, + "weighted_rkd_loss": 0.004370349110104144 + }, + { + "epoch": 8.257956448911223, + "flow_matching_loss": 0.015233697462826967, + "grad_norm": 0.18681037425994873, + "learning_rate": 1.6535246748837772e-08, + "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": 0.004151923372410238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019706888357177376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01027453546412289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015233697462826967, + "weighted_rkd_loss": 0.005137267732061445 + }, + { + "epoch": 8.274706867671691, + "flow_matching_loss": 0.017054551746696234, + "grad_norm": 0.15061014890670776, + "learning_rate": 1.2206033304855702e-08, + "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": 0.0037409895565360786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016018183785490691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00908379745669663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.017054551746696234, + "weighted_rkd_loss": 0.004541898728348315 + }, + { + "epoch": 8.291457286432161, + "flow_matching_loss": 0.014949669688940048, + "grad_norm": 0.13207487761974335, + "learning_rate": 8.532434522695031e-09, + "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": 0.004302018298767507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017844451940618455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010388481616973876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014949669688940048, + "weighted_rkd_loss": 0.005194240808486938 + }, + { + "epoch": 8.308207705192629, + "flow_matching_loss": 0.01610209411010146, + "grad_norm": 0.14128775894641876, + "learning_rate": 5.5146110974679805e-09, + "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": 0.0036398351658135653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001506704674102366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00878637502901256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.01610209411010146, + "weighted_rkd_loss": 0.00439318751450628 + }, + { + "epoch": 8.324958123953099, + "flow_matching_loss": 0.016640785429626704, + "grad_norm": 0.14564873278141022, + "learning_rate": 3.1526950385496467e-09, + "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": 0.003576566115953028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014642269699834288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00861735912039876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.016640785429626704, + "weighted_rkd_loss": 0.00430867956019938 + }, + { + "epoch": 8.341708542713567, + "flow_matching_loss": 0.013791149761527777, + "grad_norm": 0.139508917927742, + "learning_rate": 1.4467896637959666e-09, + "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": 0.0038056798744946717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014783959602937102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009089755592867732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.013791149761527777, + "weighted_rkd_loss": 0.004544877796433866 + }, + { + "epoch": 8.358458961474037, + "flow_matching_loss": 0.015681551396846773, + "grad_norm": 0.11605437844991684, + "learning_rate": 3.9696959503898733e-10, + "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": 0.004223085520789027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017177864559926093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010163957532495261, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.015681551396846773, + "weighted_rkd_loss": 0.0050819787662476305 + }, + { + "epoch": 8.375209380234505, + "flow_matching_loss": 0.014806061424314976, + "grad_norm": 0.19967901706695557, + "learning_rate": 3.280754806156416e-12, + "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": 0.004259938187897205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020176144782453774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010537490900605916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "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": 0.014806061424314976, + "weighted_rkd_loss": 0.005268745450302958 + }, + { + "epoch": 8.375209380234505, + "step": 5000, + "total_flos": 0.0, + "train_loss": 0.04256536388993263, + "train_runtime": 8905.1869, + "train_samples_per_second": 35.934, + "train_steps_per_second": 0.561 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": true + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..23dcfa871a88e56280f945357aef0296bd071711 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.577927505176127, + -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-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a3758b2dac9cab1274647a0a3e4647bb197ec7fc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6f702e66e76c9c2a5773af5ec6b03a2aabb911ab03f04ce0b6ef1d2a8dca66a +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..f0ff0c407e5f8330be930fc06e511099b9b0966b --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7c3e7fa3ab1a1b699fc070a0037b929ffb343acd8c22a8ed9fed100096b7c6cc +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/optimizer.pt b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..0860dfad6cd8aecf43cb1e079f427d5f77dd613b --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7ec5a2417760ad581d41002e8fde42354e0a457587a5baf10237ab81d92fe14a +size 1624159435 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..11634d0c8f4194799e6e92809754cb4a1d041e2d --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-1000/trainer_state.json @@ -0,0 +1,4634 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.7094017094017095, + "eval_steps": 500, + "global_step": 1000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.017094017094017096, + "flow_matching_loss": 0.20608119964599608, + "grad_norm": 1.4477208852767944, + "learning_rate": 1.08e-06, + "loss": 0.4085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1441098630428314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11659819930791855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.4048179268836975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10, + "weighted_flow_matching_loss": 0.20608119964599608, + "weighted_rkd_loss": 0.20240896344184875 + }, + { + "epoch": 0.03418803418803419, + "flow_matching_loss": 0.22261241227388381, + "grad_norm": 1.1392409801483154, + "learning_rate": 2.28e-06, + "loss": 0.4131, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13591156229376794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10920900553464889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.38103213608264924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20, + "weighted_flow_matching_loss": 0.22261241227388381, + "weighted_rkd_loss": 0.19051606804132462 + }, + { + "epoch": 0.05128205128205128, + "flow_matching_loss": 0.20390308946371077, + "grad_norm": 0.9336247444152832, + "learning_rate": 3.48e-06, + "loss": 0.388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1328342378139496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1025419719517231, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3682104468345642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 30, + "weighted_flow_matching_loss": 0.20390308946371077, + "weighted_rkd_loss": 0.1841052234172821 + }, + { + "epoch": 0.06837606837606838, + "flow_matching_loss": 0.2105909526348114, + "grad_norm": 0.8749905824661255, + "learning_rate": 4.68e-06, + "loss": 0.3894, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12913469299674035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09934469908475876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.35761408507823944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 40, + "weighted_flow_matching_loss": 0.2105909526348114, + "weighted_rkd_loss": 0.17880704253911972 + }, + { + "epoch": 0.08547008547008547, + "flow_matching_loss": 0.1939498320221901, + "grad_norm": 0.8234826326370239, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3703, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1269106835126877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09883174449205398, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3526531159877777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 50, + "weighted_flow_matching_loss": 0.1939498320221901, + "weighted_rkd_loss": 0.17632655799388885 + }, + { + "epoch": 0.10256410256410256, + "flow_matching_loss": 0.15859663859009743, + "grad_norm": 0.7916533946990967, + "learning_rate": 7.08e-06, + "loss": 0.3244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12063853591680526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09041309282183647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3316901683807373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 60, + "weighted_flow_matching_loss": 0.15859663859009743, + "weighted_rkd_loss": 0.16584508419036864 + }, + { + "epoch": 0.11965811965811966, + "flow_matching_loss": 0.15783419758081435, + "grad_norm": 0.8823524117469788, + "learning_rate": 8.28e-06, + "loss": 0.3089, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11185461878776551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0784363217651844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30214556455612185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 70, + "weighted_flow_matching_loss": 0.15783419758081435, + "weighted_rkd_loss": 0.15107278227806092 + }, + { + "epoch": 0.13675213675213677, + "flow_matching_loss": 0.13894206359982492, + "grad_norm": 0.941081702709198, + "learning_rate": 9.48e-06, + "loss": 0.2816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1060655303299427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0731036301702261, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.28523469567298887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 80, + "weighted_flow_matching_loss": 0.13894206359982492, + "weighted_rkd_loss": 0.14261734783649443 + }, + { + "epoch": 0.15384615384615385, + "flow_matching_loss": 0.12293280363082885, + "grad_norm": 0.8303114771842957, + "learning_rate": 1.068e-05, + "loss": 0.258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10080260336399079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06849895156919956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.270104156434536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 90, + "weighted_flow_matching_loss": 0.12293280363082885, + "weighted_rkd_loss": 0.135052078217268 + }, + { + "epoch": 0.17094017094017094, + "flow_matching_loss": 0.1046357847750187, + "grad_norm": 0.7039380669593811, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.2292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09367859661579132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06181581430137158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24917300790548325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 100, + "weighted_flow_matching_loss": 0.1046357847750187, + "weighted_rkd_loss": 0.12458650395274162 + }, + { + "epoch": 0.18803418803418803, + "flow_matching_loss": 0.09870888739824295, + "grad_norm": 0.6019426584243774, + "learning_rate": 1.308e-05, + "loss": 0.2142, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08723139390349388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05651620887219906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23097899854183196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 110, + "weighted_flow_matching_loss": 0.09870888739824295, + "weighted_rkd_loss": 0.11548949927091598 + }, + { + "epoch": 0.20512820512820512, + "flow_matching_loss": 0.08241323307156563, + "grad_norm": 0.6689212918281555, + "learning_rate": 1.428e-05, + "loss": 0.197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08734692260622978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0545398335903883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2292336791753769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 120, + "weighted_flow_matching_loss": 0.08241323307156563, + "weighted_rkd_loss": 0.11461683958768845 + }, + { + "epoch": 0.2222222222222222, + "flow_matching_loss": 0.0766731396317482, + "grad_norm": 0.578585684299469, + "learning_rate": 1.548e-05, + "loss": 0.1806, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07932209149003029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.049220692738890647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20786487609148024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 130, + "weighted_flow_matching_loss": 0.0766731396317482, + "weighted_rkd_loss": 0.10393243804574012 + }, + { + "epoch": 0.23931623931623933, + "flow_matching_loss": 0.08649998903274536, + "grad_norm": 0.6135573983192444, + "learning_rate": 1.6680000000000003e-05, + "loss": 0.1817, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07297333218157291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0444162305444479, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1903628960251808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 140, + "weighted_flow_matching_loss": 0.08649998903274536, + "weighted_rkd_loss": 0.0951814480125904 + }, + { + "epoch": 0.2564102564102564, + "flow_matching_loss": 0.07530920952558517, + "grad_norm": 0.6404624581336975, + "learning_rate": 1.7879999999999998e-05, + "loss": 0.1619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06675466485321521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.039724042639136314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17323337346315384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 150, + "weighted_flow_matching_loss": 0.07530920952558517, + "weighted_rkd_loss": 0.08661668673157692 + }, + { + "epoch": 0.27350427350427353, + "flow_matching_loss": 0.07482291832566261, + "grad_norm": 0.614500880241394, + "learning_rate": 1.908e-05, + "loss": 0.1624, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0673632200807333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.040419120155274865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17514555901288986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 160, + "weighted_flow_matching_loss": 0.07482291832566261, + "weighted_rkd_loss": 0.08757277950644493 + }, + { + "epoch": 0.2905982905982906, + "flow_matching_loss": 0.07167641781270503, + "grad_norm": 0.5213647484779358, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0648277897387743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03784646335989237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16750204414129258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 170, + "weighted_flow_matching_loss": 0.07167641781270503, + "weighted_rkd_loss": 0.08375102207064629 + }, + { + "epoch": 0.3076923076923077, + "flow_matching_loss": 0.06723086908459663, + "grad_norm": 0.5122416019439697, + "learning_rate": 2.148e-05, + "loss": 0.144, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05934748239815235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034743427857756616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15343839228153228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 180, + "weighted_flow_matching_loss": 0.06723086908459663, + "weighted_rkd_loss": 0.07671919614076614 + }, + { + "epoch": 0.3247863247863248, + "flow_matching_loss": 0.06515696458518505, + "grad_norm": 0.5083960890769958, + "learning_rate": 2.268e-05, + "loss": 0.1374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.055695648118853566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033016927726566794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14440822303295137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 190, + "weighted_flow_matching_loss": 0.06515696458518505, + "weighted_rkd_loss": 0.07220411151647568 + }, + { + "epoch": 0.3418803418803419, + "flow_matching_loss": 0.06426175646483898, + "grad_norm": 0.5611753463745117, + "learning_rate": 2.3880000000000002e-05, + "loss": 0.1338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.053841910883784296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03130413573235273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13898795917630197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 200, + "weighted_flow_matching_loss": 0.06426175646483898, + "weighted_rkd_loss": 0.06949397958815098 + }, + { + "epoch": 0.358974358974359, + "flow_matching_loss": 0.061613532900810244, + "grad_norm": 0.4719548523426056, + "learning_rate": 2.508e-05, + "loss": 0.1246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.048743893206119535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02855243906378746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1260402247309685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 210, + "weighted_flow_matching_loss": 0.061613532900810244, + "weighted_rkd_loss": 0.06302011236548424 + }, + { + "epoch": 0.37606837606837606, + "flow_matching_loss": 0.058244922384619716, + "grad_norm": 0.489758163690567, + "learning_rate": 2.628e-05, + "loss": 0.1197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04813622012734413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026654975302517415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12292741760611534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 220, + "weighted_flow_matching_loss": 0.058244922384619716, + "weighted_rkd_loss": 0.06146370880305767 + }, + { + "epoch": 0.39316239316239315, + "flow_matching_loss": 0.05867415480315685, + "grad_norm": 0.5043800473213196, + "learning_rate": 2.748e-05, + "loss": 0.1202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04817534200847149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02667076177895069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1230214461684227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 230, + "weighted_flow_matching_loss": 0.05867415480315685, + "weighted_rkd_loss": 0.06151072308421135 + }, + { + "epoch": 0.41025641025641024, + "flow_matching_loss": 0.059197704493999484, + "grad_norm": 0.4292409420013428, + "learning_rate": 2.868e-05, + "loss": 0.1177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04579513482749462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025458005629479884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11704827472567558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 240, + "weighted_flow_matching_loss": 0.059197704493999484, + "weighted_rkd_loss": 0.05852413736283779 + }, + { + "epoch": 0.42735042735042733, + "flow_matching_loss": 0.05531383752822876, + "grad_norm": 0.43816468119621277, + "learning_rate": 2.9880000000000002e-05, + "loss": 0.1148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.046195313706994055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026640130765736104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1190307579934597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 250, + "weighted_flow_matching_loss": 0.05531383752822876, + "weighted_rkd_loss": 0.05951537899672985 + }, + { + "epoch": 0.4444444444444444, + "flow_matching_loss": 0.05735364258289337, + "grad_norm": 0.4437594413757324, + "learning_rate": 2.9999734259635676e-05, + "loss": 0.1043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03679828383028507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020236621610820293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09383318945765495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 260, + "weighted_flow_matching_loss": 0.05735364258289337, + "weighted_rkd_loss": 0.046916594728827477 + }, + { + "epoch": 0.46153846153846156, + "flow_matching_loss": 0.059609442576766015, + "grad_norm": 0.5054996013641357, + "learning_rate": 2.9998815663057245e-05, + "loss": 0.1117, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041301059536635876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021624139230698346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10422625914216041, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 270, + "weighted_flow_matching_loss": 0.059609442576766015, + "weighted_rkd_loss": 0.052113129571080205 + }, + { + "epoch": 0.47863247863247865, + "flow_matching_loss": 0.05288516990840435, + "grad_norm": 0.48650580644607544, + "learning_rate": 2.999724096969228e-05, + "loss": 0.1047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04097450710833073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021585627086460592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10353464037179946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 280, + "weighted_flow_matching_loss": 0.05288516990840435, + "weighted_rkd_loss": 0.05176732018589973 + }, + { + "epoch": 0.49572649572649574, + "flow_matching_loss": 0.0536977294832468, + "grad_norm": 0.42119359970092773, + "learning_rate": 2.9995010248422984e-05, + "loss": 0.1018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.038251850940287116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01963985590264201, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09614355638623237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 290, + "weighted_flow_matching_loss": 0.0536977294832468, + "weighted_rkd_loss": 0.048071778193116185 + }, + { + "epoch": 0.5128205128205128, + "flow_matching_loss": 0.048235632851719855, + "grad_norm": 0.4279158115386963, + "learning_rate": 2.999212359682832e-05, + "loss": 0.1011, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04124627523124218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023284840025007726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10577739253640175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 300, + "weighted_flow_matching_loss": 0.048235632851719855, + "weighted_rkd_loss": 0.05288869626820088 + }, + { + "epoch": 0.5299145299145299, + "flow_matching_loss": 0.04915821589529514, + "grad_norm": 0.4583510756492615, + "learning_rate": 2.9988581141179787e-05, + "loss": 0.0909, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03272451851516962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018078654445707797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08352769166231155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 310, + "weighted_flow_matching_loss": 0.04915821589529514, + "weighted_rkd_loss": 0.04176384583115578 + }, + { + "epoch": 0.5470085470085471, + "flow_matching_loss": 0.0478827927261591, + "grad_norm": 0.4557395875453949, + "learning_rate": 2.998438303643587e-05, + "loss": 0.0943, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03672756850719452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01931789694353938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09277303442358971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 320, + "weighted_flow_matching_loss": 0.0478827927261591, + "weighted_rkd_loss": 0.046386517211794855 + }, + { + "epoch": 0.5641025641025641, + "flow_matching_loss": 0.048558805510401724, + "grad_norm": 0.4305928349494934, + "learning_rate": 2.997952946623527e-05, + "loss": 0.0936, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03552434463053942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019004741962999106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09005343131721019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 330, + "weighted_flow_matching_loss": 0.048558805510401724, + "weighted_rkd_loss": 0.045026715658605096 + }, + { + "epoch": 0.5811965811965812, + "flow_matching_loss": 0.05098775625228882, + "grad_norm": 0.40920335054397583, + "learning_rate": 2.9974020642888875e-05, + "loss": 0.0918, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03249767124652862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01662399461492896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08161933720111847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 340, + "weighted_flow_matching_loss": 0.05098775625228882, + "weighted_rkd_loss": 0.040809668600559235 + }, + { + "epoch": 0.5982905982905983, + "flow_matching_loss": 0.047236426174640654, + "grad_norm": 0.4020651578903198, + "learning_rate": 2.9967856807370458e-05, + "loss": 0.0876, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032064920477569106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016578875295817853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08070871606469154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 350, + "weighted_flow_matching_loss": 0.047236426174640654, + "weighted_rkd_loss": 0.04035435803234577 + }, + { + "epoch": 0.6153846153846154, + "flow_matching_loss": 0.049100768938660624, + "grad_norm": 0.410824179649353, + "learning_rate": 2.996103822930615e-05, + "loss": 0.0887, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03171112034469843, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015830404963344334, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07925264574587346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 360, + "weighted_flow_matching_loss": 0.049100768938660624, + "weighted_rkd_loss": 0.03962632287293673 + }, + { + "epoch": 0.6324786324786325, + "flow_matching_loss": 0.04570407420396805, + "grad_norm": 0.41649773716926575, + "learning_rate": 2.9953565206962653e-05, + "loss": 0.0868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03249792978167534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017130458448082208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08212631791830063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 370, + "weighted_flow_matching_loss": 0.04570407420396805, + "weighted_rkd_loss": 0.04106315895915032 + }, + { + "epoch": 0.6495726495726496, + "flow_matching_loss": 0.04507583752274513, + "grad_norm": 0.41268017888069153, + "learning_rate": 2.9945438067234172e-05, + "loss": 0.0817, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029481365904212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014191387593746186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07315411977469921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 380, + "weighted_flow_matching_loss": 0.04507583752274513, + "weighted_rkd_loss": 0.036577059887349606 + }, + { + "epoch": 0.6666666666666666, + "flow_matching_loss": 0.04255103617906571, + "grad_norm": 0.38128671050071716, + "learning_rate": 2.9936657165628124e-05, + "loss": 0.0831, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032352514937520026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016295951325446367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08100098073482513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 390, + "weighted_flow_matching_loss": 0.04255103617906571, + "weighted_rkd_loss": 0.040500490367412566 + }, + { + "epoch": 0.6837606837606838, + "flow_matching_loss": 0.04447304848581553, + "grad_norm": 0.36115458607673645, + "learning_rate": 2.992722288624959e-05, + "loss": 0.0812, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02881138101220131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01579931229352951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07342207506299019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 400, + "weighted_flow_matching_loss": 0.04447304848581553, + "weighted_rkd_loss": 0.03671103753149509 + }, + { + "epoch": 0.7008547008547008, + "flow_matching_loss": 0.04228375982493162, + "grad_norm": 0.27871495485305786, + "learning_rate": 2.9917135641784515e-05, + "loss": 0.0782, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028570003621280194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014606839325278998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07174684777855873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 410, + "weighted_flow_matching_loss": 0.04228375982493162, + "weighted_rkd_loss": 0.03587342388927937 + }, + { + "epoch": 0.717948717948718, + "flow_matching_loss": 0.044394811242818834, + "grad_norm": 0.43788713216781616, + "learning_rate": 2.9906395873481655e-05, + "loss": 0.0769, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0254867784678936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013939346186816692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06491290293633938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 420, + "weighted_flow_matching_loss": 0.044394811242818834, + "weighted_rkd_loss": 0.03245645146816969 + }, + { + "epoch": 0.7350427350427351, + "flow_matching_loss": 0.04304563030600548, + "grad_norm": 0.32673171162605286, + "learning_rate": 2.989500405113326e-05, + "loss": 0.0757, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026353750936686994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012691938411444425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06539944037795067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 430, + "weighted_flow_matching_loss": 0.04304563030600548, + "weighted_rkd_loss": 0.03269972018897534 + }, + { + "epoch": 0.7521367521367521, + "flow_matching_loss": 0.04123639427125454, + "grad_norm": 0.33855384588241577, + "learning_rate": 2.9882960673054547e-05, + "loss": 0.0736, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02597996760159731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012860519997775554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06482045613229274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 440, + "weighted_flow_matching_loss": 0.04123639427125454, + "weighted_rkd_loss": 0.03241022806614637 + }, + { + "epoch": 0.7692307692307693, + "flow_matching_loss": 0.04132717214524746, + "grad_norm": 0.3718223571777344, + "learning_rate": 2.9870266266061886e-05, + "loss": 0.0739, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026512700691819192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012095902115106583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06512130312621593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 450, + "weighted_flow_matching_loss": 0.04132717214524746, + "weighted_rkd_loss": 0.032560651563107966 + }, + { + "epoch": 0.7863247863247863, + "flow_matching_loss": 0.042681030184030536, + "grad_norm": 0.3572709560394287, + "learning_rate": 2.9856921385449766e-05, + "loss": 0.0774, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02780984025448561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013740665093064309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06936034522950649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 460, + "weighted_flow_matching_loss": 0.042681030184030536, + "weighted_rkd_loss": 0.034680172614753244 + }, + { + "epoch": 0.8034188034188035, + "flow_matching_loss": 0.042522648349404335, + "grad_norm": 0.3548585772514343, + "learning_rate": 2.9842926614966502e-05, + "loss": 0.0715, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023204243928194045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011549634858965874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057958122715353964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 470, + "weighted_flow_matching_loss": 0.042522648349404335, + "weighted_rkd_loss": 0.028979061357676982 + }, + { + "epoch": 0.8205128205128205, + "flow_matching_loss": 0.03533885534852743, + "grad_norm": 0.3358217775821686, + "learning_rate": 2.982828256678868e-05, + "loss": 0.0665, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024509145691990853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01339802285656333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06241631470620632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 480, + "weighted_flow_matching_loss": 0.03533885534852743, + "weighted_rkd_loss": 0.03120815735310316 + }, + { + "epoch": 0.8376068376068376, + "flow_matching_loss": 0.0407237846404314, + "grad_norm": 0.40146124362945557, + "learning_rate": 2.9812989881494412e-05, + "loss": 0.0751, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027627827040851115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013586688507348298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06884234212338924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 490, + "weighted_flow_matching_loss": 0.0407237846404314, + "weighted_rkd_loss": 0.03442117106169462 + }, + { + "epoch": 0.8547008547008547, + "flow_matching_loss": 0.042021133750677106, + "grad_norm": 0.2975722849369049, + "learning_rate": 2.9797049228035296e-05, + "loss": 0.0724, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02429118975996971, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01214608708396554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060728466510772704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 500, + "weighted_flow_matching_loss": 0.042021133750677106, + "weighted_rkd_loss": 0.030364233255386352 + }, + { + "epoch": 0.8717948717948718, + "flow_matching_loss": 0.03959287479519844, + "grad_norm": 0.37908297777175903, + "learning_rate": 2.978046130370715e-05, + "loss": 0.0687, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02314175870269537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011852209083735942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0581357266753912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 510, + "weighted_flow_matching_loss": 0.03959287479519844, + "weighted_rkd_loss": 0.0290678633376956 + }, + { + "epoch": 0.8888888888888888, + "flow_matching_loss": 0.038183963485062124, + "grad_norm": 0.35484832525253296, + "learning_rate": 2.9763226834119535e-05, + "loss": 0.0666, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022654735110700132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01152317225933075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05683264210820198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 520, + "weighted_flow_matching_loss": 0.038183963485062124, + "weighted_rkd_loss": 0.02841632105410099 + }, + { + "epoch": 0.905982905982906, + "flow_matching_loss": 0.03838554341346025, + "grad_norm": 0.3829350471496582, + "learning_rate": 2.9745346573163963e-05, + "loss": 0.0678, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02331932317465544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012097928486764432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.058736575022339824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 530, + "weighted_flow_matching_loss": 0.03838554341346025, + "weighted_rkd_loss": 0.029368287511169912 + }, + { + "epoch": 0.9230769230769231, + "flow_matching_loss": 0.0373695008456707, + "grad_norm": 0.39486923813819885, + "learning_rate": 2.9726821302980987e-05, + "loss": 0.0647, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02153742704540491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011604637745767832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054679491370916364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 540, + "weighted_flow_matching_loss": 0.0373695008456707, + "weighted_rkd_loss": 0.027339745685458182 + }, + { + "epoch": 0.9401709401709402, + "flow_matching_loss": 0.03420814126729965, + "grad_norm": 0.30264589190483093, + "learning_rate": 2.9707651833925935e-05, + "loss": 0.0593, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020150486007332803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009803551714867353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050104524567723276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 550, + "weighted_flow_matching_loss": 0.03420814126729965, + "weighted_rkd_loss": 0.025052262283861638 + }, + { + "epoch": 0.9572649572649573, + "flow_matching_loss": 0.03828359097242355, + "grad_norm": 0.27153193950653076, + "learning_rate": 2.968783900453348e-05, + "loss": 0.0632, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02007452305406332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009622522396966815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049771569296717645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 560, + "weighted_flow_matching_loss": 0.03828359097242355, + "weighted_rkd_loss": 0.024885784648358823 + }, + { + "epoch": 0.9743589743589743, + "flow_matching_loss": 0.03746143244206905, + "grad_norm": 0.28983309864997864, + "learning_rate": 2.9667383681480968e-05, + "loss": 0.0648, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02210656674578786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010438397945836186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05465153120458126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 570, + "weighted_flow_matching_loss": 0.03746143244206905, + "weighted_rkd_loss": 0.02732576560229063 + }, + { + "epoch": 0.9914529914529915, + "flow_matching_loss": 0.035928660072386265, + "grad_norm": 0.30345749855041504, + "learning_rate": 2.9646286759550493e-05, + "loss": 0.0628, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021595945302397012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010564278997480869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05375616922974587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 580, + "weighted_flow_matching_loss": 0.035928660072386265, + "weighted_rkd_loss": 0.026878084614872934 + }, + { + "epoch": 1.0085470085470085, + "flow_matching_loss": 0.03552918881177902, + "grad_norm": 0.24723964929580688, + "learning_rate": 2.9624549161589764e-05, + "loss": 0.0581, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018401246704161166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008261607168242336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0450641006231308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 590, + "weighted_flow_matching_loss": 0.03552918881177902, + "weighted_rkd_loss": 0.0225320503115654 + }, + { + "epoch": 1.0256410256410255, + "flow_matching_loss": 0.03407776262611151, + "grad_norm": 0.30071568489074707, + "learning_rate": 2.9602171838471745e-05, + "loss": 0.0579, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019089525938034056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009438153170049191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04761720485985279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 600, + "weighted_flow_matching_loss": 0.03407776262611151, + "weighted_rkd_loss": 0.023808602429926395 + }, + { + "epoch": 1.0427350427350428, + "flow_matching_loss": 0.03366336897015572, + "grad_norm": 0.302155464887619, + "learning_rate": 2.957915576905303e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02271198071539402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011890822742134332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057314783707261084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 610, + "weighted_flow_matching_loss": 0.03366336897015572, + "weighted_rkd_loss": 0.028657391853630542 + }, + { + "epoch": 1.0598290598290598, + "flow_matching_loss": 0.03334424737840891, + "grad_norm": 0.31046783924102783, + "learning_rate": 2.9555501960131058e-05, + "loss": 0.0577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019524640496820213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009662961680442094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04871224202215672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 620, + "weighted_flow_matching_loss": 0.03334424737840891, + "weighted_rkd_loss": 0.02435612101107836 + }, + { + "epoch": 1.0769230769230769, + "flow_matching_loss": 0.03312153853476048, + "grad_norm": 0.28929588198661804, + "learning_rate": 2.9531211446400068e-05, + "loss": 0.0552, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017922693397849798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00835406300611794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044199449568986894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 630, + "weighted_flow_matching_loss": 0.03312153853476048, + "weighted_rkd_loss": 0.022099724784493447 + }, + { + "epoch": 1.0940170940170941, + "flow_matching_loss": 0.031021049618721007, + "grad_norm": 0.29626306891441345, + "learning_rate": 2.9506285290405813e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018444733694195747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008001519087702035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0448909867554903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 640, + "weighted_flow_matching_loss": 0.031021049618721007, + "weighted_rkd_loss": 0.02244549337774515 + }, + { + "epoch": 1.1111111111111112, + "flow_matching_loss": 0.032624239847064015, + "grad_norm": 0.2606738209724426, + "learning_rate": 2.948072458249911e-05, + "loss": 0.0572, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01954005965963006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010045465733855962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04912558533251286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 650, + "weighted_flow_matching_loss": 0.032624239847064015, + "weighted_rkd_loss": 0.02456279266625643 + }, + { + "epoch": 1.1282051282051282, + "flow_matching_loss": 0.0357012577354908, + "grad_norm": 0.35832640528678894, + "learning_rate": 2.9454530440788125e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01891506966203451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008951532281935215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04678167216479778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 660, + "weighted_flow_matching_loss": 0.0357012577354908, + "weighted_rkd_loss": 0.02339083608239889 + }, + { + "epoch": 1.1452991452991452, + "flow_matching_loss": 0.035934435203671455, + "grad_norm": 0.3307911455631256, + "learning_rate": 2.9427704011089474e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018776980880647897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008833648543804884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046387610211968425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 670, + "weighted_flow_matching_loss": 0.035934435203671455, + "weighted_rkd_loss": 0.023193805105984212 + }, + { + "epoch": 1.1623931623931625, + "flow_matching_loss": 0.035431265085935595, + "grad_norm": 0.32725879549980164, + "learning_rate": 2.94002464668781e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018233825359493493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008717586379498243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045185237750411035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 680, + "weighted_flow_matching_loss": 0.035431265085935595, + "weighted_rkd_loss": 0.022592618875205518 + }, + { + "epoch": 1.1794871794871795, + "flow_matching_loss": 0.03349699769169092, + "grad_norm": 0.3060956299304962, + "learning_rate": 2.9372159009235936e-05, + "loss": 0.0549, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017201956920325757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008407466765493154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042811380326747896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 690, + "weighted_flow_matching_loss": 0.03349699769169092, + "weighted_rkd_loss": 0.021405690163373948 + }, + { + "epoch": 1.1965811965811965, + "flow_matching_loss": 0.037000064924359324, + "grad_norm": 0.2392890304327011, + "learning_rate": 2.9343442866799365e-05, + "loss": 0.0589, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01774967508390546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008203081460669637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04370243176817894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 700, + "weighted_flow_matching_loss": 0.037000064924359324, + "weighted_rkd_loss": 0.02185121588408947 + }, + { + "epoch": 1.2136752136752136, + "flow_matching_loss": 0.0321138696745038, + "grad_norm": 0.2744523286819458, + "learning_rate": 2.9314099295705495e-05, + "loss": 0.0545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01787513755261898, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008945518126711249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04469579309225082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 710, + "weighted_flow_matching_loss": 0.0321138696745038, + "weighted_rkd_loss": 0.02234789654612541 + }, + { + "epoch": 1.2307692307692308, + "flow_matching_loss": 0.03291467633098364, + "grad_norm": 0.27376607060432434, + "learning_rate": 2.928412957953718e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0181625971570611, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008992884447798133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04531807880848646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 720, + "weighted_flow_matching_loss": 0.03291467633098364, + "weighted_rkd_loss": 0.02265903940424323 + }, + { + "epoch": 1.2478632478632479, + "flow_matching_loss": 0.031597768515348436, + "grad_norm": 0.29815053939819336, + "learning_rate": 2.925353502926689e-05, + "loss": 0.0513, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015860845521092414, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007615697477012873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03933738861232996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 730, + "weighted_flow_matching_loss": 0.031597768515348436, + "weighted_rkd_loss": 0.01966869430616498 + }, + { + "epoch": 1.264957264957265, + "flow_matching_loss": 0.03209205660969019, + "grad_norm": 0.29948389530181885, + "learning_rate": 2.922231698319937e-05, + "loss": 0.0538, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017416194640100004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008499196963384748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04333158619701862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 740, + "weighted_flow_matching_loss": 0.03209205660969019, + "weighted_rkd_loss": 0.02166579309850931 + }, + { + "epoch": 1.282051282051282, + "flow_matching_loss": 0.02969713993370533, + "grad_norm": 0.2717636227607727, + "learning_rate": 2.9190476806913088e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0162487150169909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007376636797562242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039874066784977916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 750, + "weighted_flow_matching_loss": 0.02969713993370533, + "weighted_rkd_loss": 0.019937033392488958 + }, + { + "epoch": 1.2991452991452992, + "flow_matching_loss": 0.031202525086700916, + "grad_norm": 0.28742673993110657, + "learning_rate": 2.9158015893200504e-05, + "loss": 0.0521, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017281018663197756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007276421086862684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04183845836669207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 760, + "weighted_flow_matching_loss": 0.031202525086700916, + "weighted_rkd_loss": 0.020919229183346034 + }, + { + "epoch": 1.3162393162393162, + "flow_matching_loss": 0.03131227232515812, + "grad_norm": 0.28647977113723755, + "learning_rate": 2.9124935662007137e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01633641654625535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007625138806179166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040297971293330195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 770, + "weighted_flow_matching_loss": 0.03131227232515812, + "weighted_rkd_loss": 0.020148985646665098 + }, + { + "epoch": 1.3333333333333333, + "flow_matching_loss": 0.031031945906579494, + "grad_norm": 0.3045690655708313, + "learning_rate": 2.9091237560369467e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01676619527861476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007660567946732044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041192957758903505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 780, + "weighted_flow_matching_loss": 0.031031945906579494, + "weighted_rkd_loss": 0.020596478879451752 + }, + { + "epoch": 1.3504273504273505, + "flow_matching_loss": 0.03271263670176268, + "grad_norm": 0.2736513912677765, + "learning_rate": 2.905692306235163e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01463465103879571, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00730584766715765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03657514955848455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 790, + "weighted_flow_matching_loss": 0.03271263670176268, + "weighted_rkd_loss": 0.018287574779242276 + }, + { + "epoch": 1.3675213675213675, + "flow_matching_loss": 0.030691403523087503, + "grad_norm": 0.28391724824905396, + "learning_rate": 2.9021993668980925e-05, + "loss": 0.0507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016008828580379487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008079428132623434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040097085386514665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 800, + "weighted_flow_matching_loss": 0.030691403523087503, + "weighted_rkd_loss": 0.020048542693257333 + }, + { + "epoch": 1.3846153846153846, + "flow_matching_loss": 0.034423727355897424, + "grad_norm": 0.27780401706695557, + "learning_rate": 2.8986450908182185e-05, + "loss": 0.0546, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016429053619503975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007567548984661698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04042565673589706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 810, + "weighted_flow_matching_loss": 0.034423727355897424, + "weighted_rkd_loss": 0.02021282836794853 + }, + { + "epoch": 1.4017094017094016, + "flow_matching_loss": 0.03095700442790985, + "grad_norm": 0.22609981894493103, + "learning_rate": 2.8950296334710902e-05, + "loss": 0.0519, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01725160414353013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007418723125010729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041921931132674214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 820, + "weighted_flow_matching_loss": 0.03095700442790985, + "weighted_rkd_loss": 0.020960965566337107 + }, + { + "epoch": 1.4188034188034189, + "flow_matching_loss": 0.03410516288131475, + "grad_norm": 0.2484390139579773, + "learning_rate": 2.8913531530085248e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015002989396452903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007663191389292478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037669170275330545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 830, + "weighted_flow_matching_loss": 0.03410516288131475, + "weighted_rkd_loss": 0.018834585137665272 + }, + { + "epoch": 1.435897435897436, + "flow_matching_loss": 0.030957960151135923, + "grad_norm": 0.2720315754413605, + "learning_rate": 2.887615810251687e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016805268358439207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007622529938817024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04123306684195995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 840, + "weighted_flow_matching_loss": 0.030957960151135923, + "weighted_rkd_loss": 0.020616533420979976 + }, + { + "epoch": 1.452991452991453, + "flow_matching_loss": 0.030326111428439616, + "grad_norm": 0.2526715099811554, + "learning_rate": 2.8838177686840577e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014052478969097138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006443384615704417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03454834185540676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 850, + "weighted_flow_matching_loss": 0.030326111428439616, + "weighted_rkd_loss": 0.01727417092770338 + }, + { + "epoch": 1.4700854700854702, + "flow_matching_loss": 0.03168223239481449, + "grad_norm": 0.3726175129413605, + "learning_rate": 2.8799591944442776e-05, + "loss": 0.0518, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01616350244730711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007998800138011575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04032580554485321, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 860, + "weighted_flow_matching_loss": 0.03168223239481449, + "weighted_rkd_loss": 0.020162902772426605 + }, + { + "epoch": 1.4871794871794872, + "flow_matching_loss": 0.030286968126893042, + "grad_norm": 0.24847525358200073, + "learning_rate": 2.8760402563188848e-05, + "loss": 0.0486, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014799616392701864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007003285316750407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03660251796245575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 870, + "weighted_flow_matching_loss": 0.030286968126893042, + "weighted_rkd_loss": 0.018301258981227874 + }, + { + "epoch": 1.5042735042735043, + "flow_matching_loss": 0.029968923516571523, + "grad_norm": 0.31482070684432983, + "learning_rate": 2.8720611257349285e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01352565586566925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006156855309382081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03320816699415445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 880, + "weighted_flow_matching_loss": 0.029968923516571523, + "weighted_rkd_loss": 0.016604083497077225 + }, + { + "epoch": 1.5213675213675213, + "flow_matching_loss": 0.03208826966583729, + "grad_norm": 0.26682037115097046, + "learning_rate": 2.868021976752471e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012783072888851166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006290466827340424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031856612861156465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 890, + "weighted_flow_matching_loss": 0.03208826966583729, + "weighted_rkd_loss": 0.015928306430578233 + }, + { + "epoch": 1.5384615384615383, + "flow_matching_loss": 0.029574750922620296, + "grad_norm": 0.2852403223514557, + "learning_rate": 2.8639229860569738e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01526008890941739, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071847605984658005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03770493865013123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 900, + "weighted_flow_matching_loss": 0.029574750922620296, + "weighted_rkd_loss": 0.018852469325065613 + }, + { + "epoch": 1.5555555555555556, + "flow_matching_loss": 0.028749799728393553, + "grad_norm": 0.24580393731594086, + "learning_rate": 2.8597643329515688e-05, + "loss": 0.0447, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012961071636527776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006020331801846623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031942475400865075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 910, + "weighted_flow_matching_loss": 0.028749799728393553, + "weighted_rkd_loss": 0.015971237700432538 + }, + { + "epoch": 1.5726495726495726, + "flow_matching_loss": 0.032204396836459634, + "grad_norm": 0.20449334383010864, + "learning_rate": 2.855546199349214e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014906360022723674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006444963859394193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0362576836720109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 920, + "weighted_flow_matching_loss": 0.032204396836459634, + "weighted_rkd_loss": 0.01812884183600545 + }, + { + "epoch": 1.5897435897435899, + "flow_matching_loss": 0.029012378305196762, + "grad_norm": 0.2885199785232544, + "learning_rate": 2.85126876976474e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015990046598017216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007635087100788951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03961518015712499, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 930, + "weighted_flow_matching_loss": 0.029012378305196762, + "weighted_rkd_loss": 0.019807590078562497 + }, + { + "epoch": 1.606837606837607, + "flow_matching_loss": 0.0270179970189929, + "grad_norm": 0.2731475830078125, + "learning_rate": 2.8469322313067726e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014710118900984526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006910082930698991, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03633032031357288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 940, + "weighted_flow_matching_loss": 0.0270179970189929, + "weighted_rkd_loss": 0.01816516015678644 + }, + { + "epoch": 1.623931623931624, + "flow_matching_loss": 0.03037530407309532, + "grad_norm": 0.19273969531059265, + "learning_rate": 2.8425367736695522e-05, + "loss": 0.0482, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014435206167399883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00685276158619672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03572317436337471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 950, + "weighted_flow_matching_loss": 0.03037530407309532, + "weighted_rkd_loss": 0.017861587181687354 + }, + { + "epoch": 1.641025641025641, + "flow_matching_loss": 0.027571728639304637, + "grad_norm": 0.25559166073799133, + "learning_rate": 2.838082589124636e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012530108634382486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006172595894895494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031232813000679018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 960, + "weighted_flow_matching_loss": 0.027571728639304637, + "weighted_rkd_loss": 0.015616406500339509 + }, + { + "epoch": 1.658119658119658, + "flow_matching_loss": 0.026423130929470063, + "grad_norm": 0.2502352297306061, + "learning_rate": 2.8335698725124843e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013723570108413696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00633350177668035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0337806424126029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 970, + "weighted_flow_matching_loss": 0.026423130929470063, + "weighted_rkd_loss": 0.01689032120630145 + }, + { + "epoch": 1.6752136752136753, + "flow_matching_loss": 0.026968703232705594, + "grad_norm": 0.29313844442367554, + "learning_rate": 2.8289988212339422e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015268613491207362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006759832962416112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0372970599681139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 980, + "weighted_flow_matching_loss": 0.026968703232705594, + "weighted_rkd_loss": 0.01864852998405695 + }, + { + "epoch": 1.6923076923076923, + "flow_matching_loss": 0.025319616310298443, + "grad_norm": 0.2522859573364258, + "learning_rate": 2.8243696352415997e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014617102593183518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006836280273273587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03607048541307449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 990, + "weighted_flow_matching_loss": 0.025319616310298443, + "weighted_rkd_loss": 0.018035242706537245 + }, + { + "epoch": 1.7094017094017095, + "flow_matching_loss": 0.030002601444721222, + "grad_norm": 0.24149076640605927, + "learning_rate": 2.8196825170310495e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01380759608000517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006553543242625892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03416873514652252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1000, + "weighted_flow_matching_loss": 0.030002601444721222, + "weighted_rkd_loss": 0.01708436757326126 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..23dcfa871a88e56280f945357aef0296bd071711 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.577927505176127, + -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-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..01178ef3c09cc6ce273bff419795986a4f29e0db --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:573166cad39b1b44d5bba8ea1adf4538562996cd229b54d1ddc9950212288f3c +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..b4a9650ebdab72789d0c5ea5ec01b9630157cc0d --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-3000/trainer_state.json @@ -0,0 +1,13834 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 5.128205128205128, + "eval_steps": 500, + "global_step": 3000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.017094017094017096, + "flow_matching_loss": 0.20608119964599608, + "grad_norm": 1.4477208852767944, + "learning_rate": 1.08e-06, + "loss": 0.4085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1441098630428314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11659819930791855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.4048179268836975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10, + "weighted_flow_matching_loss": 0.20608119964599608, + "weighted_rkd_loss": 0.20240896344184875 + }, + { + "epoch": 0.03418803418803419, + "flow_matching_loss": 0.22261241227388381, + "grad_norm": 1.1392409801483154, + "learning_rate": 2.28e-06, + "loss": 0.4131, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13591156229376794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10920900553464889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.38103213608264924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20, + "weighted_flow_matching_loss": 0.22261241227388381, + "weighted_rkd_loss": 0.19051606804132462 + }, + { + "epoch": 0.05128205128205128, + "flow_matching_loss": 0.20390308946371077, + "grad_norm": 0.9336247444152832, + "learning_rate": 3.48e-06, + "loss": 0.388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1328342378139496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1025419719517231, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3682104468345642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 30, + "weighted_flow_matching_loss": 0.20390308946371077, + "weighted_rkd_loss": 0.1841052234172821 + }, + { + "epoch": 0.06837606837606838, + "flow_matching_loss": 0.2105909526348114, + "grad_norm": 0.8749905824661255, + "learning_rate": 4.68e-06, + "loss": 0.3894, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12913469299674035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09934469908475876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.35761408507823944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 40, + "weighted_flow_matching_loss": 0.2105909526348114, + "weighted_rkd_loss": 0.17880704253911972 + }, + { + "epoch": 0.08547008547008547, + "flow_matching_loss": 0.1939498320221901, + "grad_norm": 0.8234826326370239, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3703, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1269106835126877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09883174449205398, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3526531159877777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 50, + "weighted_flow_matching_loss": 0.1939498320221901, + "weighted_rkd_loss": 0.17632655799388885 + }, + { + "epoch": 0.10256410256410256, + "flow_matching_loss": 0.15859663859009743, + "grad_norm": 0.7916533946990967, + "learning_rate": 7.08e-06, + "loss": 0.3244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12063853591680526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09041309282183647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3316901683807373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 60, + "weighted_flow_matching_loss": 0.15859663859009743, + "weighted_rkd_loss": 0.16584508419036864 + }, + { + "epoch": 0.11965811965811966, + "flow_matching_loss": 0.15783419758081435, + "grad_norm": 0.8823524117469788, + "learning_rate": 8.28e-06, + "loss": 0.3089, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11185461878776551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0784363217651844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30214556455612185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 70, + "weighted_flow_matching_loss": 0.15783419758081435, + "weighted_rkd_loss": 0.15107278227806092 + }, + { + "epoch": 0.13675213675213677, + "flow_matching_loss": 0.13894206359982492, + "grad_norm": 0.941081702709198, + "learning_rate": 9.48e-06, + "loss": 0.2816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1060655303299427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0731036301702261, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.28523469567298887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 80, + "weighted_flow_matching_loss": 0.13894206359982492, + "weighted_rkd_loss": 0.14261734783649443 + }, + { + "epoch": 0.15384615384615385, + "flow_matching_loss": 0.12293280363082885, + "grad_norm": 0.8303114771842957, + "learning_rate": 1.068e-05, + "loss": 0.258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10080260336399079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06849895156919956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.270104156434536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 90, + "weighted_flow_matching_loss": 0.12293280363082885, + "weighted_rkd_loss": 0.135052078217268 + }, + { + "epoch": 0.17094017094017094, + "flow_matching_loss": 0.1046357847750187, + "grad_norm": 0.7039380669593811, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.2292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09367859661579132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06181581430137158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24917300790548325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 100, + "weighted_flow_matching_loss": 0.1046357847750187, + "weighted_rkd_loss": 0.12458650395274162 + }, + { + "epoch": 0.18803418803418803, + "flow_matching_loss": 0.09870888739824295, + "grad_norm": 0.6019426584243774, + "learning_rate": 1.308e-05, + "loss": 0.2142, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08723139390349388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05651620887219906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23097899854183196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 110, + "weighted_flow_matching_loss": 0.09870888739824295, + "weighted_rkd_loss": 0.11548949927091598 + }, + { + "epoch": 0.20512820512820512, + "flow_matching_loss": 0.08241323307156563, + "grad_norm": 0.6689212918281555, + "learning_rate": 1.428e-05, + "loss": 0.197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08734692260622978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0545398335903883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2292336791753769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 120, + "weighted_flow_matching_loss": 0.08241323307156563, + "weighted_rkd_loss": 0.11461683958768845 + }, + { + "epoch": 0.2222222222222222, + "flow_matching_loss": 0.0766731396317482, + "grad_norm": 0.578585684299469, + "learning_rate": 1.548e-05, + "loss": 0.1806, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07932209149003029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.049220692738890647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20786487609148024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 130, + "weighted_flow_matching_loss": 0.0766731396317482, + "weighted_rkd_loss": 0.10393243804574012 + }, + { + "epoch": 0.23931623931623933, + "flow_matching_loss": 0.08649998903274536, + "grad_norm": 0.6135573983192444, + "learning_rate": 1.6680000000000003e-05, + "loss": 0.1817, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07297333218157291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0444162305444479, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1903628960251808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 140, + "weighted_flow_matching_loss": 0.08649998903274536, + "weighted_rkd_loss": 0.0951814480125904 + }, + { + "epoch": 0.2564102564102564, + "flow_matching_loss": 0.07530920952558517, + "grad_norm": 0.6404624581336975, + "learning_rate": 1.7879999999999998e-05, + "loss": 0.1619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06675466485321521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.039724042639136314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17323337346315384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 150, + "weighted_flow_matching_loss": 0.07530920952558517, + "weighted_rkd_loss": 0.08661668673157692 + }, + { + "epoch": 0.27350427350427353, + "flow_matching_loss": 0.07482291832566261, + "grad_norm": 0.614500880241394, + "learning_rate": 1.908e-05, + "loss": 0.1624, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0673632200807333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.040419120155274865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17514555901288986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 160, + "weighted_flow_matching_loss": 0.07482291832566261, + "weighted_rkd_loss": 0.08757277950644493 + }, + { + "epoch": 0.2905982905982906, + "flow_matching_loss": 0.07167641781270503, + "grad_norm": 0.5213647484779358, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0648277897387743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03784646335989237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16750204414129258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 170, + "weighted_flow_matching_loss": 0.07167641781270503, + "weighted_rkd_loss": 0.08375102207064629 + }, + { + "epoch": 0.3076923076923077, + "flow_matching_loss": 0.06723086908459663, + "grad_norm": 0.5122416019439697, + "learning_rate": 2.148e-05, + "loss": 0.144, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05934748239815235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034743427857756616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15343839228153228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 180, + "weighted_flow_matching_loss": 0.06723086908459663, + "weighted_rkd_loss": 0.07671919614076614 + }, + { + "epoch": 0.3247863247863248, + "flow_matching_loss": 0.06515696458518505, + "grad_norm": 0.5083960890769958, + "learning_rate": 2.268e-05, + "loss": 0.1374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.055695648118853566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033016927726566794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14440822303295137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 190, + "weighted_flow_matching_loss": 0.06515696458518505, + "weighted_rkd_loss": 0.07220411151647568 + }, + { + "epoch": 0.3418803418803419, + "flow_matching_loss": 0.06426175646483898, + "grad_norm": 0.5611753463745117, + "learning_rate": 2.3880000000000002e-05, + "loss": 0.1338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.053841910883784296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03130413573235273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13898795917630197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 200, + "weighted_flow_matching_loss": 0.06426175646483898, + "weighted_rkd_loss": 0.06949397958815098 + }, + { + "epoch": 0.358974358974359, + "flow_matching_loss": 0.061613532900810244, + "grad_norm": 0.4719548523426056, + "learning_rate": 2.508e-05, + "loss": 0.1246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.048743893206119535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02855243906378746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1260402247309685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 210, + "weighted_flow_matching_loss": 0.061613532900810244, + "weighted_rkd_loss": 0.06302011236548424 + }, + { + "epoch": 0.37606837606837606, + "flow_matching_loss": 0.058244922384619716, + "grad_norm": 0.489758163690567, + "learning_rate": 2.628e-05, + "loss": 0.1197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04813622012734413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026654975302517415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12292741760611534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 220, + "weighted_flow_matching_loss": 0.058244922384619716, + "weighted_rkd_loss": 0.06146370880305767 + }, + { + "epoch": 0.39316239316239315, + "flow_matching_loss": 0.05867415480315685, + "grad_norm": 0.5043800473213196, + "learning_rate": 2.748e-05, + "loss": 0.1202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04817534200847149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02667076177895069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1230214461684227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 230, + "weighted_flow_matching_loss": 0.05867415480315685, + "weighted_rkd_loss": 0.06151072308421135 + }, + { + "epoch": 0.41025641025641024, + "flow_matching_loss": 0.059197704493999484, + "grad_norm": 0.4292409420013428, + "learning_rate": 2.868e-05, + "loss": 0.1177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04579513482749462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025458005629479884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11704827472567558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 240, + "weighted_flow_matching_loss": 0.059197704493999484, + "weighted_rkd_loss": 0.05852413736283779 + }, + { + "epoch": 0.42735042735042733, + "flow_matching_loss": 0.05531383752822876, + "grad_norm": 0.43816468119621277, + "learning_rate": 2.9880000000000002e-05, + "loss": 0.1148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.046195313706994055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026640130765736104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1190307579934597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 250, + "weighted_flow_matching_loss": 0.05531383752822876, + "weighted_rkd_loss": 0.05951537899672985 + }, + { + "epoch": 0.4444444444444444, + "flow_matching_loss": 0.05735364258289337, + "grad_norm": 0.4437594413757324, + "learning_rate": 2.9999734259635676e-05, + "loss": 0.1043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03679828383028507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020236621610820293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09383318945765495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 260, + "weighted_flow_matching_loss": 0.05735364258289337, + "weighted_rkd_loss": 0.046916594728827477 + }, + { + "epoch": 0.46153846153846156, + "flow_matching_loss": 0.059609442576766015, + "grad_norm": 0.5054996013641357, + "learning_rate": 2.9998815663057245e-05, + "loss": 0.1117, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041301059536635876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021624139230698346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10422625914216041, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 270, + "weighted_flow_matching_loss": 0.059609442576766015, + "weighted_rkd_loss": 0.052113129571080205 + }, + { + "epoch": 0.47863247863247865, + "flow_matching_loss": 0.05288516990840435, + "grad_norm": 0.48650580644607544, + "learning_rate": 2.999724096969228e-05, + "loss": 0.1047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04097450710833073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021585627086460592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10353464037179946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 280, + "weighted_flow_matching_loss": 0.05288516990840435, + "weighted_rkd_loss": 0.05176732018589973 + }, + { + "epoch": 0.49572649572649574, + "flow_matching_loss": 0.0536977294832468, + "grad_norm": 0.42119359970092773, + "learning_rate": 2.9995010248422984e-05, + "loss": 0.1018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.038251850940287116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01963985590264201, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09614355638623237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 290, + "weighted_flow_matching_loss": 0.0536977294832468, + "weighted_rkd_loss": 0.048071778193116185 + }, + { + "epoch": 0.5128205128205128, + "flow_matching_loss": 0.048235632851719855, + "grad_norm": 0.4279158115386963, + "learning_rate": 2.999212359682832e-05, + "loss": 0.1011, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04124627523124218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023284840025007726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10577739253640175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 300, + "weighted_flow_matching_loss": 0.048235632851719855, + "weighted_rkd_loss": 0.05288869626820088 + }, + { + "epoch": 0.5299145299145299, + "flow_matching_loss": 0.04915821589529514, + "grad_norm": 0.4583510756492615, + "learning_rate": 2.9988581141179787e-05, + "loss": 0.0909, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03272451851516962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018078654445707797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08352769166231155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 310, + "weighted_flow_matching_loss": 0.04915821589529514, + "weighted_rkd_loss": 0.04176384583115578 + }, + { + "epoch": 0.5470085470085471, + "flow_matching_loss": 0.0478827927261591, + "grad_norm": 0.4557395875453949, + "learning_rate": 2.998438303643587e-05, + "loss": 0.0943, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03672756850719452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01931789694353938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09277303442358971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 320, + "weighted_flow_matching_loss": 0.0478827927261591, + "weighted_rkd_loss": 0.046386517211794855 + }, + { + "epoch": 0.5641025641025641, + "flow_matching_loss": 0.048558805510401724, + "grad_norm": 0.4305928349494934, + "learning_rate": 2.997952946623527e-05, + "loss": 0.0936, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03552434463053942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019004741962999106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09005343131721019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 330, + "weighted_flow_matching_loss": 0.048558805510401724, + "weighted_rkd_loss": 0.045026715658605096 + }, + { + "epoch": 0.5811965811965812, + "flow_matching_loss": 0.05098775625228882, + "grad_norm": 0.40920335054397583, + "learning_rate": 2.9974020642888875e-05, + "loss": 0.0918, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03249767124652862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01662399461492896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08161933720111847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 340, + "weighted_flow_matching_loss": 0.05098775625228882, + "weighted_rkd_loss": 0.040809668600559235 + }, + { + "epoch": 0.5982905982905983, + "flow_matching_loss": 0.047236426174640654, + "grad_norm": 0.4020651578903198, + "learning_rate": 2.9967856807370458e-05, + "loss": 0.0876, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032064920477569106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016578875295817853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08070871606469154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 350, + "weighted_flow_matching_loss": 0.047236426174640654, + "weighted_rkd_loss": 0.04035435803234577 + }, + { + "epoch": 0.6153846153846154, + "flow_matching_loss": 0.049100768938660624, + "grad_norm": 0.410824179649353, + "learning_rate": 2.996103822930615e-05, + "loss": 0.0887, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03171112034469843, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015830404963344334, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07925264574587346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 360, + "weighted_flow_matching_loss": 0.049100768938660624, + "weighted_rkd_loss": 0.03962632287293673 + }, + { + "epoch": 0.6324786324786325, + "flow_matching_loss": 0.04570407420396805, + "grad_norm": 0.41649773716926575, + "learning_rate": 2.9953565206962653e-05, + "loss": 0.0868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03249792978167534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017130458448082208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08212631791830063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 370, + "weighted_flow_matching_loss": 0.04570407420396805, + "weighted_rkd_loss": 0.04106315895915032 + }, + { + "epoch": 0.6495726495726496, + "flow_matching_loss": 0.04507583752274513, + "grad_norm": 0.41268017888069153, + "learning_rate": 2.9945438067234172e-05, + "loss": 0.0817, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029481365904212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014191387593746186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07315411977469921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 380, + "weighted_flow_matching_loss": 0.04507583752274513, + "weighted_rkd_loss": 0.036577059887349606 + }, + { + "epoch": 0.6666666666666666, + "flow_matching_loss": 0.04255103617906571, + "grad_norm": 0.38128671050071716, + "learning_rate": 2.9936657165628124e-05, + "loss": 0.0831, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032352514937520026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016295951325446367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08100098073482513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 390, + "weighted_flow_matching_loss": 0.04255103617906571, + "weighted_rkd_loss": 0.040500490367412566 + }, + { + "epoch": 0.6837606837606838, + "flow_matching_loss": 0.04447304848581553, + "grad_norm": 0.36115458607673645, + "learning_rate": 2.992722288624959e-05, + "loss": 0.0812, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02881138101220131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01579931229352951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07342207506299019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 400, + "weighted_flow_matching_loss": 0.04447304848581553, + "weighted_rkd_loss": 0.03671103753149509 + }, + { + "epoch": 0.7008547008547008, + "flow_matching_loss": 0.04228375982493162, + "grad_norm": 0.27871495485305786, + "learning_rate": 2.9917135641784515e-05, + "loss": 0.0782, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028570003621280194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014606839325278998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07174684777855873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 410, + "weighted_flow_matching_loss": 0.04228375982493162, + "weighted_rkd_loss": 0.03587342388927937 + }, + { + "epoch": 0.717948717948718, + "flow_matching_loss": 0.044394811242818834, + "grad_norm": 0.43788713216781616, + "learning_rate": 2.9906395873481655e-05, + "loss": 0.0769, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0254867784678936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013939346186816692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06491290293633938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 420, + "weighted_flow_matching_loss": 0.044394811242818834, + "weighted_rkd_loss": 0.03245645146816969 + }, + { + "epoch": 0.7350427350427351, + "flow_matching_loss": 0.04304563030600548, + "grad_norm": 0.32673171162605286, + "learning_rate": 2.989500405113326e-05, + "loss": 0.0757, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026353750936686994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012691938411444425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06539944037795067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 430, + "weighted_flow_matching_loss": 0.04304563030600548, + "weighted_rkd_loss": 0.03269972018897534 + }, + { + "epoch": 0.7521367521367521, + "flow_matching_loss": 0.04123639427125454, + "grad_norm": 0.33855384588241577, + "learning_rate": 2.9882960673054547e-05, + "loss": 0.0736, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02597996760159731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012860519997775554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06482045613229274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 440, + "weighted_flow_matching_loss": 0.04123639427125454, + "weighted_rkd_loss": 0.03241022806614637 + }, + { + "epoch": 0.7692307692307693, + "flow_matching_loss": 0.04132717214524746, + "grad_norm": 0.3718223571777344, + "learning_rate": 2.9870266266061886e-05, + "loss": 0.0739, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026512700691819192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012095902115106583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06512130312621593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 450, + "weighted_flow_matching_loss": 0.04132717214524746, + "weighted_rkd_loss": 0.032560651563107966 + }, + { + "epoch": 0.7863247863247863, + "flow_matching_loss": 0.042681030184030536, + "grad_norm": 0.3572709560394287, + "learning_rate": 2.9856921385449766e-05, + "loss": 0.0774, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02780984025448561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013740665093064309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06936034522950649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 460, + "weighted_flow_matching_loss": 0.042681030184030536, + "weighted_rkd_loss": 0.034680172614753244 + }, + { + "epoch": 0.8034188034188035, + "flow_matching_loss": 0.042522648349404335, + "grad_norm": 0.3548585772514343, + "learning_rate": 2.9842926614966502e-05, + "loss": 0.0715, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023204243928194045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011549634858965874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057958122715353964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 470, + "weighted_flow_matching_loss": 0.042522648349404335, + "weighted_rkd_loss": 0.028979061357676982 + }, + { + "epoch": 0.8205128205128205, + "flow_matching_loss": 0.03533885534852743, + "grad_norm": 0.3358217775821686, + "learning_rate": 2.982828256678868e-05, + "loss": 0.0665, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024509145691990853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01339802285656333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06241631470620632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 480, + "weighted_flow_matching_loss": 0.03533885534852743, + "weighted_rkd_loss": 0.03120815735310316 + }, + { + "epoch": 0.8376068376068376, + "flow_matching_loss": 0.0407237846404314, + "grad_norm": 0.40146124362945557, + "learning_rate": 2.9812989881494412e-05, + "loss": 0.0751, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027627827040851115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013586688507348298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06884234212338924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 490, + "weighted_flow_matching_loss": 0.0407237846404314, + "weighted_rkd_loss": 0.03442117106169462 + }, + { + "epoch": 0.8547008547008547, + "flow_matching_loss": 0.042021133750677106, + "grad_norm": 0.2975722849369049, + "learning_rate": 2.9797049228035296e-05, + "loss": 0.0724, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02429118975996971, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01214608708396554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060728466510772704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 500, + "weighted_flow_matching_loss": 0.042021133750677106, + "weighted_rkd_loss": 0.030364233255386352 + }, + { + "epoch": 0.8717948717948718, + "flow_matching_loss": 0.03959287479519844, + "grad_norm": 0.37908297777175903, + "learning_rate": 2.978046130370715e-05, + "loss": 0.0687, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02314175870269537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011852209083735942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0581357266753912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 510, + "weighted_flow_matching_loss": 0.03959287479519844, + "weighted_rkd_loss": 0.0290678633376956 + }, + { + "epoch": 0.8888888888888888, + "flow_matching_loss": 0.038183963485062124, + "grad_norm": 0.35484832525253296, + "learning_rate": 2.9763226834119535e-05, + "loss": 0.0666, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022654735110700132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01152317225933075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05683264210820198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 520, + "weighted_flow_matching_loss": 0.038183963485062124, + "weighted_rkd_loss": 0.02841632105410099 + }, + { + "epoch": 0.905982905982906, + "flow_matching_loss": 0.03838554341346025, + "grad_norm": 0.3829350471496582, + "learning_rate": 2.9745346573163963e-05, + "loss": 0.0678, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02331932317465544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012097928486764432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.058736575022339824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 530, + "weighted_flow_matching_loss": 0.03838554341346025, + "weighted_rkd_loss": 0.029368287511169912 + }, + { + "epoch": 0.9230769230769231, + "flow_matching_loss": 0.0373695008456707, + "grad_norm": 0.39486923813819885, + "learning_rate": 2.9726821302980987e-05, + "loss": 0.0647, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02153742704540491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011604637745767832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054679491370916364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 540, + "weighted_flow_matching_loss": 0.0373695008456707, + "weighted_rkd_loss": 0.027339745685458182 + }, + { + "epoch": 0.9401709401709402, + "flow_matching_loss": 0.03420814126729965, + "grad_norm": 0.30264589190483093, + "learning_rate": 2.9707651833925935e-05, + "loss": 0.0593, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020150486007332803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009803551714867353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050104524567723276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 550, + "weighted_flow_matching_loss": 0.03420814126729965, + "weighted_rkd_loss": 0.025052262283861638 + }, + { + "epoch": 0.9572649572649573, + "flow_matching_loss": 0.03828359097242355, + "grad_norm": 0.27153193950653076, + "learning_rate": 2.968783900453348e-05, + "loss": 0.0632, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02007452305406332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009622522396966815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049771569296717645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 560, + "weighted_flow_matching_loss": 0.03828359097242355, + "weighted_rkd_loss": 0.024885784648358823 + }, + { + "epoch": 0.9743589743589743, + "flow_matching_loss": 0.03746143244206905, + "grad_norm": 0.28983309864997864, + "learning_rate": 2.9667383681480968e-05, + "loss": 0.0648, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02210656674578786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010438397945836186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05465153120458126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 570, + "weighted_flow_matching_loss": 0.03746143244206905, + "weighted_rkd_loss": 0.02732576560229063 + }, + { + "epoch": 0.9914529914529915, + "flow_matching_loss": 0.035928660072386265, + "grad_norm": 0.30345749855041504, + "learning_rate": 2.9646286759550493e-05, + "loss": 0.0628, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021595945302397012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010564278997480869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05375616922974587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 580, + "weighted_flow_matching_loss": 0.035928660072386265, + "weighted_rkd_loss": 0.026878084614872934 + }, + { + "epoch": 1.0085470085470085, + "flow_matching_loss": 0.03552918881177902, + "grad_norm": 0.24723964929580688, + "learning_rate": 2.9624549161589764e-05, + "loss": 0.0581, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018401246704161166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008261607168242336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0450641006231308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 590, + "weighted_flow_matching_loss": 0.03552918881177902, + "weighted_rkd_loss": 0.0225320503115654 + }, + { + "epoch": 1.0256410256410255, + "flow_matching_loss": 0.03407776262611151, + "grad_norm": 0.30071568489074707, + "learning_rate": 2.9602171838471745e-05, + "loss": 0.0579, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019089525938034056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009438153170049191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04761720485985279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 600, + "weighted_flow_matching_loss": 0.03407776262611151, + "weighted_rkd_loss": 0.023808602429926395 + }, + { + "epoch": 1.0427350427350428, + "flow_matching_loss": 0.03366336897015572, + "grad_norm": 0.302155464887619, + "learning_rate": 2.957915576905303e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02271198071539402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011890822742134332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057314783707261084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 610, + "weighted_flow_matching_loss": 0.03366336897015572, + "weighted_rkd_loss": 0.028657391853630542 + }, + { + "epoch": 1.0598290598290598, + "flow_matching_loss": 0.03334424737840891, + "grad_norm": 0.31046783924102783, + "learning_rate": 2.9555501960131058e-05, + "loss": 0.0577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019524640496820213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009662961680442094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04871224202215672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 620, + "weighted_flow_matching_loss": 0.03334424737840891, + "weighted_rkd_loss": 0.02435612101107836 + }, + { + "epoch": 1.0769230769230769, + "flow_matching_loss": 0.03312153853476048, + "grad_norm": 0.28929588198661804, + "learning_rate": 2.9531211446400068e-05, + "loss": 0.0552, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017922693397849798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00835406300611794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044199449568986894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 630, + "weighted_flow_matching_loss": 0.03312153853476048, + "weighted_rkd_loss": 0.022099724784493447 + }, + { + "epoch": 1.0940170940170941, + "flow_matching_loss": 0.031021049618721007, + "grad_norm": 0.29626306891441345, + "learning_rate": 2.9506285290405813e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018444733694195747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008001519087702035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0448909867554903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 640, + "weighted_flow_matching_loss": 0.031021049618721007, + "weighted_rkd_loss": 0.02244549337774515 + }, + { + "epoch": 1.1111111111111112, + "flow_matching_loss": 0.032624239847064015, + "grad_norm": 0.2606738209724426, + "learning_rate": 2.948072458249911e-05, + "loss": 0.0572, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01954005965963006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010045465733855962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04912558533251286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 650, + "weighted_flow_matching_loss": 0.032624239847064015, + "weighted_rkd_loss": 0.02456279266625643 + }, + { + "epoch": 1.1282051282051282, + "flow_matching_loss": 0.0357012577354908, + "grad_norm": 0.35832640528678894, + "learning_rate": 2.9454530440788125e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01891506966203451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008951532281935215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04678167216479778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 660, + "weighted_flow_matching_loss": 0.0357012577354908, + "weighted_rkd_loss": 0.02339083608239889 + }, + { + "epoch": 1.1452991452991452, + "flow_matching_loss": 0.035934435203671455, + "grad_norm": 0.3307911455631256, + "learning_rate": 2.9427704011089474e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018776980880647897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008833648543804884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046387610211968425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 670, + "weighted_flow_matching_loss": 0.035934435203671455, + "weighted_rkd_loss": 0.023193805105984212 + }, + { + "epoch": 1.1623931623931625, + "flow_matching_loss": 0.035431265085935595, + "grad_norm": 0.32725879549980164, + "learning_rate": 2.94002464668781e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018233825359493493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008717586379498243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045185237750411035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 680, + "weighted_flow_matching_loss": 0.035431265085935595, + "weighted_rkd_loss": 0.022592618875205518 + }, + { + "epoch": 1.1794871794871795, + "flow_matching_loss": 0.03349699769169092, + "grad_norm": 0.3060956299304962, + "learning_rate": 2.9372159009235936e-05, + "loss": 0.0549, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017201956920325757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008407466765493154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042811380326747896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 690, + "weighted_flow_matching_loss": 0.03349699769169092, + "weighted_rkd_loss": 0.021405690163373948 + }, + { + "epoch": 1.1965811965811965, + "flow_matching_loss": 0.037000064924359324, + "grad_norm": 0.2392890304327011, + "learning_rate": 2.9343442866799365e-05, + "loss": 0.0589, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01774967508390546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008203081460669637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04370243176817894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 700, + "weighted_flow_matching_loss": 0.037000064924359324, + "weighted_rkd_loss": 0.02185121588408947 + }, + { + "epoch": 1.2136752136752136, + "flow_matching_loss": 0.0321138696745038, + "grad_norm": 0.2744523286819458, + "learning_rate": 2.9314099295705495e-05, + "loss": 0.0545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01787513755261898, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008945518126711249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04469579309225082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 710, + "weighted_flow_matching_loss": 0.0321138696745038, + "weighted_rkd_loss": 0.02234789654612541 + }, + { + "epoch": 1.2307692307692308, + "flow_matching_loss": 0.03291467633098364, + "grad_norm": 0.27376607060432434, + "learning_rate": 2.928412957953718e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0181625971570611, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008992884447798133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04531807880848646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 720, + "weighted_flow_matching_loss": 0.03291467633098364, + "weighted_rkd_loss": 0.02265903940424323 + }, + { + "epoch": 1.2478632478632479, + "flow_matching_loss": 0.031597768515348436, + "grad_norm": 0.29815053939819336, + "learning_rate": 2.925353502926689e-05, + "loss": 0.0513, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015860845521092414, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007615697477012873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03933738861232996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 730, + "weighted_flow_matching_loss": 0.031597768515348436, + "weighted_rkd_loss": 0.01966869430616498 + }, + { + "epoch": 1.264957264957265, + "flow_matching_loss": 0.03209205660969019, + "grad_norm": 0.29948389530181885, + "learning_rate": 2.922231698319937e-05, + "loss": 0.0538, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017416194640100004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008499196963384748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04333158619701862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 740, + "weighted_flow_matching_loss": 0.03209205660969019, + "weighted_rkd_loss": 0.02166579309850931 + }, + { + "epoch": 1.282051282051282, + "flow_matching_loss": 0.02969713993370533, + "grad_norm": 0.2717636227607727, + "learning_rate": 2.9190476806913088e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0162487150169909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007376636797562242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039874066784977916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 750, + "weighted_flow_matching_loss": 0.02969713993370533, + "weighted_rkd_loss": 0.019937033392488958 + }, + { + "epoch": 1.2991452991452992, + "flow_matching_loss": 0.031202525086700916, + "grad_norm": 0.28742673993110657, + "learning_rate": 2.9158015893200504e-05, + "loss": 0.0521, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017281018663197756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007276421086862684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04183845836669207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 760, + "weighted_flow_matching_loss": 0.031202525086700916, + "weighted_rkd_loss": 0.020919229183346034 + }, + { + "epoch": 1.3162393162393162, + "flow_matching_loss": 0.03131227232515812, + "grad_norm": 0.28647977113723755, + "learning_rate": 2.9124935662007137e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01633641654625535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007625138806179166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040297971293330195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 770, + "weighted_flow_matching_loss": 0.03131227232515812, + "weighted_rkd_loss": 0.020148985646665098 + }, + { + "epoch": 1.3333333333333333, + "flow_matching_loss": 0.031031945906579494, + "grad_norm": 0.3045690655708313, + "learning_rate": 2.9091237560369467e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01676619527861476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007660567946732044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041192957758903505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 780, + "weighted_flow_matching_loss": 0.031031945906579494, + "weighted_rkd_loss": 0.020596478879451752 + }, + { + "epoch": 1.3504273504273505, + "flow_matching_loss": 0.03271263670176268, + "grad_norm": 0.2736513912677765, + "learning_rate": 2.905692306235163e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01463465103879571, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00730584766715765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03657514955848455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 790, + "weighted_flow_matching_loss": 0.03271263670176268, + "weighted_rkd_loss": 0.018287574779242276 + }, + { + "epoch": 1.3675213675213675, + "flow_matching_loss": 0.030691403523087503, + "grad_norm": 0.28391724824905396, + "learning_rate": 2.9021993668980925e-05, + "loss": 0.0507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016008828580379487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008079428132623434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040097085386514665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 800, + "weighted_flow_matching_loss": 0.030691403523087503, + "weighted_rkd_loss": 0.020048542693257333 + }, + { + "epoch": 1.3846153846153846, + "flow_matching_loss": 0.034423727355897424, + "grad_norm": 0.27780401706695557, + "learning_rate": 2.8986450908182185e-05, + "loss": 0.0546, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016429053619503975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007567548984661698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04042565673589706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 810, + "weighted_flow_matching_loss": 0.034423727355897424, + "weighted_rkd_loss": 0.02021282836794853 + }, + { + "epoch": 1.4017094017094016, + "flow_matching_loss": 0.03095700442790985, + "grad_norm": 0.22609981894493103, + "learning_rate": 2.8950296334710902e-05, + "loss": 0.0519, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01725160414353013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007418723125010729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041921931132674214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 820, + "weighted_flow_matching_loss": 0.03095700442790985, + "weighted_rkd_loss": 0.020960965566337107 + }, + { + "epoch": 1.4188034188034189, + "flow_matching_loss": 0.03410516288131475, + "grad_norm": 0.2484390139579773, + "learning_rate": 2.8913531530085248e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015002989396452903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007663191389292478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037669170275330545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 830, + "weighted_flow_matching_loss": 0.03410516288131475, + "weighted_rkd_loss": 0.018834585137665272 + }, + { + "epoch": 1.435897435897436, + "flow_matching_loss": 0.030957960151135923, + "grad_norm": 0.2720315754413605, + "learning_rate": 2.887615810251687e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016805268358439207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007622529938817024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04123306684195995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 840, + "weighted_flow_matching_loss": 0.030957960151135923, + "weighted_rkd_loss": 0.020616533420979976 + }, + { + "epoch": 1.452991452991453, + "flow_matching_loss": 0.030326111428439616, + "grad_norm": 0.2526715099811554, + "learning_rate": 2.8838177686840577e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014052478969097138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006443384615704417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03454834185540676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 850, + "weighted_flow_matching_loss": 0.030326111428439616, + "weighted_rkd_loss": 0.01727417092770338 + }, + { + "epoch": 1.4700854700854702, + "flow_matching_loss": 0.03168223239481449, + "grad_norm": 0.3726175129413605, + "learning_rate": 2.8799591944442776e-05, + "loss": 0.0518, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01616350244730711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007998800138011575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04032580554485321, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 860, + "weighted_flow_matching_loss": 0.03168223239481449, + "weighted_rkd_loss": 0.020162902772426605 + }, + { + "epoch": 1.4871794871794872, + "flow_matching_loss": 0.030286968126893042, + "grad_norm": 0.24847525358200073, + "learning_rate": 2.8760402563188848e-05, + "loss": 0.0486, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014799616392701864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007003285316750407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03660251796245575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 870, + "weighted_flow_matching_loss": 0.030286968126893042, + "weighted_rkd_loss": 0.018301258981227874 + }, + { + "epoch": 1.5042735042735043, + "flow_matching_loss": 0.029968923516571523, + "grad_norm": 0.31482070684432983, + "learning_rate": 2.8720611257349285e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01352565586566925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006156855309382081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03320816699415445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 880, + "weighted_flow_matching_loss": 0.029968923516571523, + "weighted_rkd_loss": 0.016604083497077225 + }, + { + "epoch": 1.5213675213675213, + "flow_matching_loss": 0.03208826966583729, + "grad_norm": 0.26682037115097046, + "learning_rate": 2.868021976752471e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012783072888851166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006290466827340424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031856612861156465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 890, + "weighted_flow_matching_loss": 0.03208826966583729, + "weighted_rkd_loss": 0.015928306430578233 + }, + { + "epoch": 1.5384615384615383, + "flow_matching_loss": 0.029574750922620296, + "grad_norm": 0.2852403223514557, + "learning_rate": 2.8639229860569738e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01526008890941739, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071847605984658005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03770493865013123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 900, + "weighted_flow_matching_loss": 0.029574750922620296, + "weighted_rkd_loss": 0.018852469325065613 + }, + { + "epoch": 1.5555555555555556, + "flow_matching_loss": 0.028749799728393553, + "grad_norm": 0.24580393731594086, + "learning_rate": 2.8597643329515688e-05, + "loss": 0.0447, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012961071636527776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006020331801846623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031942475400865075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 910, + "weighted_flow_matching_loss": 0.028749799728393553, + "weighted_rkd_loss": 0.015971237700432538 + }, + { + "epoch": 1.5726495726495726, + "flow_matching_loss": 0.032204396836459634, + "grad_norm": 0.20449334383010864, + "learning_rate": 2.855546199349214e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014906360022723674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006444963859394193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0362576836720109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 920, + "weighted_flow_matching_loss": 0.032204396836459634, + "weighted_rkd_loss": 0.01812884183600545 + }, + { + "epoch": 1.5897435897435899, + "flow_matching_loss": 0.029012378305196762, + "grad_norm": 0.2885199785232544, + "learning_rate": 2.85126876976474e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015990046598017216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007635087100788951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03961518015712499, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 930, + "weighted_flow_matching_loss": 0.029012378305196762, + "weighted_rkd_loss": 0.019807590078562497 + }, + { + "epoch": 1.606837606837607, + "flow_matching_loss": 0.0270179970189929, + "grad_norm": 0.2731475830078125, + "learning_rate": 2.8469322313067726e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014710118900984526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006910082930698991, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03633032031357288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 940, + "weighted_flow_matching_loss": 0.0270179970189929, + "weighted_rkd_loss": 0.01816516015678644 + }, + { + "epoch": 1.623931623931624, + "flow_matching_loss": 0.03037530407309532, + "grad_norm": 0.19273969531059265, + "learning_rate": 2.8425367736695522e-05, + "loss": 0.0482, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014435206167399883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00685276158619672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03572317436337471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 950, + "weighted_flow_matching_loss": 0.03037530407309532, + "weighted_rkd_loss": 0.017861587181687354 + }, + { + "epoch": 1.641025641025641, + "flow_matching_loss": 0.027571728639304637, + "grad_norm": 0.25559166073799133, + "learning_rate": 2.838082589124636e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012530108634382486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006172595894895494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031232813000679018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 960, + "weighted_flow_matching_loss": 0.027571728639304637, + "weighted_rkd_loss": 0.015616406500339509 + }, + { + "epoch": 1.658119658119658, + "flow_matching_loss": 0.026423130929470063, + "grad_norm": 0.2502352297306061, + "learning_rate": 2.8335698725124843e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013723570108413696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00633350177668035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0337806424126029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 970, + "weighted_flow_matching_loss": 0.026423130929470063, + "weighted_rkd_loss": 0.01689032120630145 + }, + { + "epoch": 1.6752136752136753, + "flow_matching_loss": 0.026968703232705594, + "grad_norm": 0.29313844442367554, + "learning_rate": 2.8289988212339422e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015268613491207362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006759832962416112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0372970599681139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 980, + "weighted_flow_matching_loss": 0.026968703232705594, + "weighted_rkd_loss": 0.01864852998405695 + }, + { + "epoch": 1.6923076923076923, + "flow_matching_loss": 0.025319616310298443, + "grad_norm": 0.2522859573364258, + "learning_rate": 2.8243696352415997e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014617102593183518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006836280273273587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03607048541307449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 990, + "weighted_flow_matching_loss": 0.025319616310298443, + "weighted_rkd_loss": 0.018035242706537245 + }, + { + "epoch": 1.7094017094017095, + "flow_matching_loss": 0.030002601444721222, + "grad_norm": 0.24149076640605927, + "learning_rate": 2.8196825170310495e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01380759608000517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006553543242625892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03416873514652252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1000, + "weighted_flow_matching_loss": 0.030002601444721222, + "weighted_rkd_loss": 0.01708436757326126 + }, + { + "epoch": 1.7264957264957266, + "flow_matching_loss": 0.02807784080505371, + "grad_norm": 0.25452151894569397, + "learning_rate": 2.8149376716320256e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014782947115600109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00671710423193872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03628299850970507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1010, + "weighted_flow_matching_loss": 0.02807784080505371, + "weighted_rkd_loss": 0.018141499254852533 + }, + { + "epoch": 1.7435897435897436, + "flow_matching_loss": 0.02710084058344364, + "grad_norm": 0.33143165707588196, + "learning_rate": 2.8101353065994373e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014535702299326658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006557283597066999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03562868814915419, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1020, + "weighted_flow_matching_loss": 0.02710084058344364, + "weighted_rkd_loss": 0.017814344074577094 + }, + { + "epoch": 1.7606837606837606, + "flow_matching_loss": 0.02593045849353075, + "grad_norm": 0.3127861022949219, + "learning_rate": 2.8052756320042885e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01566762672737241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007534478232264519, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03886973131448031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1030, + "weighted_flow_matching_loss": 0.02593045849353075, + "weighted_rkd_loss": 0.019434865657240154 + }, + { + "epoch": 1.7777777777777777, + "flow_matching_loss": 0.0273212568834424, + "grad_norm": 0.2180718183517456, + "learning_rate": 2.8003588604244893e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01422199010848999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00723591735586524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035679897479712966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1040, + "weighted_flow_matching_loss": 0.0273212568834424, + "weighted_rkd_loss": 0.017839948739856483 + }, + { + "epoch": 1.7948717948717947, + "flow_matching_loss": 0.027072438225150108, + "grad_norm": 0.2212599217891693, + "learning_rate": 2.7953852069355565e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013318956829607487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0059395078103989364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03257742114365101, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1050, + "weighted_flow_matching_loss": 0.027072438225150108, + "weighted_rkd_loss": 0.016288710571825505 + }, + { + "epoch": 1.811965811965812, + "flow_matching_loss": 0.026676180213689803, + "grad_norm": 0.2384062260389328, + "learning_rate": 2.790354889101206e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011322647333145142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005065834778361022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027711129188537596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1060, + "weighted_flow_matching_loss": 0.026676180213689803, + "weighted_rkd_loss": 0.013855564594268798 + }, + { + "epoch": 1.8290598290598292, + "flow_matching_loss": 0.028350574523210527, + "grad_norm": 0.26893094182014465, + "learning_rate": 2.7852681269638366e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01224717302247882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005544273066334426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030038618855178355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1070, + "weighted_flow_matching_loss": 0.028350574523210527, + "weighted_rkd_loss": 0.015019309427589177 + }, + { + "epoch": 1.8461538461538463, + "flow_matching_loss": 0.023936538957059383, + "grad_norm": 0.32209914922714233, + "learning_rate": 2.780125143034902e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012277705408632756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00516571004409343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02972112111747265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1080, + "weighted_flow_matching_loss": 0.023936538957059383, + "weighted_rkd_loss": 0.014860560558736325 + }, + { + "epoch": 1.8632478632478633, + "flow_matching_loss": 0.02863725144416094, + "grad_norm": 0.22413775324821472, + "learning_rate": 2.7749261622851812e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012101378757506609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005476271221414209, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029679028689861296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1090, + "weighted_flow_matching_loss": 0.02863725144416094, + "weighted_rkd_loss": 0.014839514344930648 + }, + { + "epoch": 1.8803418803418803, + "flow_matching_loss": 0.027668011374771595, + "grad_norm": 0.26382219791412354, + "learning_rate": 2.7696714121349335e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013915220554918051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005969888484105468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03380032908171415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1100, + "weighted_flow_matching_loss": 0.027668011374771595, + "weighted_rkd_loss": 0.016900164540857077 + }, + { + "epoch": 1.8974358974358974, + "flow_matching_loss": 0.02955530472099781, + "grad_norm": 0.27939197421073914, + "learning_rate": 2.764361122443954e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01350008575245738, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00708389279898256, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034084064327180384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1110, + "weighted_flow_matching_loss": 0.02955530472099781, + "weighted_rkd_loss": 0.017042032163590192 + }, + { + "epoch": 1.9145299145299144, + "flow_matching_loss": 0.02437609266489744, + "grad_norm": 0.2628627419471741, + "learning_rate": 2.758995525501517e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01297243358567357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005478202877566218, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03142307046800852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1120, + "weighted_flow_matching_loss": 0.02437609266489744, + "weighted_rkd_loss": 0.01571153523400426 + }, + { + "epoch": 1.9316239316239316, + "flow_matching_loss": 0.028911850042641164, + "grad_norm": 0.23939956724643707, + "learning_rate": 2.753574856016215e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012397017190232873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005886134272441268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03068016842007637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1130, + "weighted_flow_matching_loss": 0.028911850042641164, + "weighted_rkd_loss": 0.015340084210038185 + }, + { + "epoch": 1.9487179487179487, + "flow_matching_loss": 0.02508879117667675, + "grad_norm": 0.20714369416236877, + "learning_rate": 2.7480993511056913e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01294971164315939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063501600408926604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03224958349019289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1140, + "weighted_flow_matching_loss": 0.02508879117667675, + "weighted_rkd_loss": 0.016124791745096446 + }, + { + "epoch": 1.965811965811966, + "flow_matching_loss": 0.024527727998793124, + "grad_norm": 0.19642457365989685, + "learning_rate": 2.7425692502862684e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011007771594449878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052364618517458435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027252004854381083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1150, + "weighted_flow_matching_loss": 0.024527727998793124, + "weighted_rkd_loss": 0.013626002427190542 + }, + { + "epoch": 1.982905982905983, + "flow_matching_loss": 0.02737492471933365, + "grad_norm": 0.23660895228385925, + "learning_rate": 2.7369847954624728e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0125096146017313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061733081936836244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031192537397146225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1160, + "weighted_flow_matching_loss": 0.02737492471933365, + "weighted_rkd_loss": 0.015596268698573112 + }, + { + "epoch": 2.0, + "flow_matching_loss": 0.026675242371857165, + "grad_norm": 0.25652265548706055, + "learning_rate": 2.731346230916448e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013060001470148564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006334699317812919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03245470225811005, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1170, + "weighted_flow_matching_loss": 0.026675242371857165, + "weighted_rkd_loss": 0.016227351129055025 + }, + { + "epoch": 2.017094017094017, + "flow_matching_loss": 0.025632075406610964, + "grad_norm": 0.22836749255657196, + "learning_rate": 2.7256538032972746e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012878391984850168, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005053568026050925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030810352042317392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1180, + "weighted_flow_matching_loss": 0.025632075406610964, + "weighted_rkd_loss": 0.015405176021158696 + }, + { + "epoch": 2.034188034188034, + "flow_matching_loss": 0.027616259641945362, + "grad_norm": 0.22967053949832916, + "learning_rate": 2.7199077616101773e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012935011554509402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005700286431238055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03157030940055847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1190, + "weighted_flow_matching_loss": 0.027616259641945362, + "weighted_rkd_loss": 0.015785154700279237 + }, + { + "epoch": 2.051282051282051, + "flow_matching_loss": 0.02580515146255493, + "grad_norm": 0.2185813933610916, + "learning_rate": 2.714108357205634e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013044429570436477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005809265840798617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03189812451601028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1200, + "weighted_flow_matching_loss": 0.02580515146255493, + "weighted_rkd_loss": 0.01594906225800514 + }, + { + "epoch": 2.0683760683760686, + "flow_matching_loss": 0.02779768891632557, + "grad_norm": 0.21726644039154053, + "learning_rate": 2.7082558437683803e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012998757604509591, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005653172452002764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03165068775415421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1210, + "weighted_flow_matching_loss": 0.02779768891632557, + "weighted_rkd_loss": 0.015825343877077104 + }, + { + "epoch": 2.0854700854700856, + "flow_matching_loss": 0.0263330465182662, + "grad_norm": 0.22308462858200073, + "learning_rate": 2.702350477306315e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011554498877376317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005302437697537243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028411435894668102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1220, + "weighted_flow_matching_loss": 0.0263330465182662, + "weighted_rkd_loss": 0.014205717947334051 + }, + { + "epoch": 2.1025641025641026, + "flow_matching_loss": 0.02595714498311281, + "grad_norm": 0.22663040459156036, + "learning_rate": 2.6963925161392963e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011306988494470715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004992392286658287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027606369741261006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1230, + "weighted_flow_matching_loss": 0.02595714498311281, + "weighted_rkd_loss": 0.013803184870630503 + }, + { + "epoch": 2.1196581196581197, + "flow_matching_loss": 0.024864679016172887, + "grad_norm": 0.3029778301715851, + "learning_rate": 2.6903822208878476e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01236183405853808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00573265366256237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030456322245299816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1240, + "weighted_flow_matching_loss": 0.024864679016172887, + "weighted_rkd_loss": 0.015228161122649908 + }, + { + "epoch": 2.1367521367521367, + "flow_matching_loss": 0.02325292509049177, + "grad_norm": 0.2267565280199051, + "learning_rate": 2.6843198544617535e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011847892682999373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00517009780742228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028865883499383925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1250, + "weighted_flow_matching_loss": 0.02325292509049177, + "weighted_rkd_loss": 0.014432941749691963 + }, + { + "epoch": 2.1538461538461537, + "flow_matching_loss": 0.02737160213291645, + "grad_norm": 0.20818153023719788, + "learning_rate": 2.6782056820485612e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012656896701082587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00568916043266654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031002953462302685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1260, + "weighted_flow_matching_loss": 0.02737160213291645, + "weighted_rkd_loss": 0.015501476731151343 + }, + { + "epoch": 2.1709401709401708, + "flow_matching_loss": 0.026486062444746494, + "grad_norm": 0.2249724268913269, + "learning_rate": 2.672039971101979e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011717427894473076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005194938997738064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028629794716835022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1270, + "weighted_flow_matching_loss": 0.026486062444746494, + "weighted_rkd_loss": 0.014314897358417511 + }, + { + "epoch": 2.1880341880341883, + "flow_matching_loss": 0.025420571491122246, + "grad_norm": 0.27121660113334656, + "learning_rate": 2.665822991330178e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012049693893641234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005436556437052787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029535944014787673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1280, + "weighted_flow_matching_loss": 0.025420571491122246, + "weighted_rkd_loss": 0.014767972007393836 + }, + { + "epoch": 2.2051282051282053, + "flow_matching_loss": 0.025054209306836127, + "grad_norm": 0.3212575316429138, + "learning_rate": 2.6595550146839934e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011498598288744688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0056119068060070275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02860910352319479, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1290, + "weighted_flow_matching_loss": 0.025054209306836127, + "weighted_rkd_loss": 0.014304551761597395 + }, + { + "epoch": 2.2222222222222223, + "flow_matching_loss": 0.02525439951568842, + "grad_norm": 0.26853254437446594, + "learning_rate": 2.6532363153450287e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012216195557266473, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005422738334164024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029855129588395358, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1300, + "weighted_flow_matching_loss": 0.02525439951568842, + "weighted_rkd_loss": 0.014927564794197679 + }, + { + "epoch": 2.2393162393162394, + "flow_matching_loss": 0.021937138680368663, + "grad_norm": 0.1946357786655426, + "learning_rate": 2.6468671697136634e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01110919527709484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005047859624028206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02726624980568886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1310, + "weighted_flow_matching_loss": 0.021937138680368663, + "weighted_rkd_loss": 0.01363312490284443 + }, + { + "epoch": 2.2564102564102564, + "flow_matching_loss": 0.02496455330401659, + "grad_norm": 0.22266526520252228, + "learning_rate": 2.640447856396961e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011567748803645372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050113315926864745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02814682889729738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1320, + "weighted_flow_matching_loss": 0.02496455330401659, + "weighted_rkd_loss": 0.01407341444864869 + }, + { + "epoch": 2.2735042735042734, + "flow_matching_loss": 0.02324515637010336, + "grad_norm": 0.22817914187908173, + "learning_rate": 2.6339786561964805e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012382819782942533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005309685179963708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03007532488554716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1330, + "weighted_flow_matching_loss": 0.02324515637010336, + "weighted_rkd_loss": 0.01503766244277358 + }, + { + "epoch": 2.2905982905982905, + "flow_matching_loss": 0.023643253184854985, + "grad_norm": 0.21733051538467407, + "learning_rate": 2.6274598520959966e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010999999195337295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004985643154941499, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026985641196370126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1340, + "weighted_flow_matching_loss": 0.023643253184854985, + "weighted_rkd_loss": 0.013492820598185063 + }, + { + "epoch": 2.3076923076923075, + "flow_matching_loss": 0.022906559705734252, + "grad_norm": 0.28345248103141785, + "learning_rate": 2.6208917292491185e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011069928156211972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005206426791846752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027346283197402954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1350, + "weighted_flow_matching_loss": 0.022906559705734252, + "weighted_rkd_loss": 0.013673141598701477 + }, + { + "epoch": 2.324786324786325, + "flow_matching_loss": 0.02266085259616375, + "grad_norm": 0.1736982762813568, + "learning_rate": 2.6142745749668176e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011071230052039027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004988636053167283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02713109627366066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1360, + "weighted_flow_matching_loss": 0.02266085259616375, + "weighted_rkd_loss": 0.01356554813683033 + }, + { + "epoch": 2.341880341880342, + "flow_matching_loss": 0.02306621465831995, + "grad_norm": 0.26756277680397034, + "learning_rate": 2.607608678704859e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010481357015669346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004531278763897717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0254939928650856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1370, + "weighted_flow_matching_loss": 0.02306621465831995, + "weighted_rkd_loss": 0.0127469964325428 + }, + { + "epoch": 2.358974358974359, + "flow_matching_loss": 0.02586138006299734, + "grad_norm": 0.24713675677776337, + "learning_rate": 2.600894332051139e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011580483987927437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004711030353792012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027871998213231563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1380, + "weighted_flow_matching_loss": 0.02586138006299734, + "weighted_rkd_loss": 0.013935999106615781 + }, + { + "epoch": 2.376068376068376, + "flow_matching_loss": 0.026022879406809807, + "grad_norm": 0.21917226910591125, + "learning_rate": 2.594131828712932e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011406512837857008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004805121943354607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027618147805333136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1390, + "weighted_flow_matching_loss": 0.026022879406809807, + "weighted_rkd_loss": 0.013809073902666568 + }, + { + "epoch": 2.393162393162393, + "flow_matching_loss": 0.023596177250146864, + "grad_norm": 0.2035856992006302, + "learning_rate": 2.5873214645040414e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010142395226284861, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045007105683907865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024785500764846802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1400, + "weighted_flow_matching_loss": 0.023596177250146864, + "weighted_rkd_loss": 0.012392750382423401 + }, + { + "epoch": 2.41025641025641, + "flow_matching_loss": 0.02610907070338726, + "grad_norm": 0.21590881049633026, + "learning_rate": 2.5804635373318604e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008909029187634587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037137356353923678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02153179394081235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1410, + "weighted_flow_matching_loss": 0.02610907070338726, + "weighted_rkd_loss": 0.010765896970406174 + }, + { + "epoch": 2.427350427350427, + "flow_matching_loss": 0.02329109571874142, + "grad_norm": 0.2922363579273224, + "learning_rate": 2.57355834718434e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011788456933572888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004883810528554022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02846072446554899, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1420, + "weighted_flow_matching_loss": 0.02329109571874142, + "weighted_rkd_loss": 0.014230362232774496 + }, + { + "epoch": 2.4444444444444446, + "flow_matching_loss": 0.024115246534347535, + "grad_norm": 0.2587732970714569, + "learning_rate": 2.566606196116867e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013214311562478542, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005839744093827903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03226836714893579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1430, + "weighted_flow_matching_loss": 0.024115246534347535, + "weighted_rkd_loss": 0.016134183574467897 + }, + { + "epoch": 2.4615384615384617, + "flow_matching_loss": 0.02489748504012823, + "grad_norm": 0.21666760742664337, + "learning_rate": 2.559607388239051e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010760731622576713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004546113940887153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026067577116191388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1440, + "weighted_flow_matching_loss": 0.02489748504012823, + "weighted_rkd_loss": 0.013033788558095694 + }, + { + "epoch": 2.4786324786324787, + "flow_matching_loss": 0.020900059305131436, + "grad_norm": 0.2407768964767456, + "learning_rate": 2.552562229701422e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010831856867298484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004969510156661272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02663322389125824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1450, + "weighted_flow_matching_loss": 0.020900059305131436, + "weighted_rkd_loss": 0.01331661194562912 + }, + { + "epoch": 2.4957264957264957, + "flow_matching_loss": 0.022048553638160228, + "grad_norm": 0.16484962403774261, + "learning_rate": 2.545471028682036e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008939298242330551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003671724908053875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02155032129958272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1460, + "weighted_flow_matching_loss": 0.022048553638160228, + "weighted_rkd_loss": 0.01077516064979136 + }, + { + "epoch": 2.5128205128205128, + "flow_matching_loss": 0.024487999454140662, + "grad_norm": 0.1850469410419464, + "learning_rate": 2.538334095373e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010523739224299788, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00469970719423145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025747185945510863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1470, + "weighted_flow_matching_loss": 0.024487999454140662, + "weighted_rkd_loss": 0.012873592972755431 + }, + { + "epoch": 2.52991452991453, + "flow_matching_loss": 0.023977626115083694, + "grad_norm": 0.26976409554481506, + "learning_rate": 2.531151741966896e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01056251721456647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005071718548424542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026196752954274417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1480, + "weighted_flow_matching_loss": 0.023977626115083694, + "weighted_rkd_loss": 0.013098376477137208 + }, + { + "epoch": 2.547008547008547, + "flow_matching_loss": 0.02404904179275036, + "grad_norm": 0.22599978744983673, + "learning_rate": 2.5239242826431302e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010430255578830839, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004550645523704588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02541115628555417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1490, + "weighted_flow_matching_loss": 0.02404904179275036, + "weighted_rkd_loss": 0.012705578142777086 + }, + { + "epoch": 2.564102564102564, + "flow_matching_loss": 0.026297237910330295, + "grad_norm": 0.2992944121360779, + "learning_rate": 2.5166520335541865e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011783476080745458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052417038241401315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02880865577608347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1500, + "weighted_flow_matching_loss": 0.026297237910330295, + "weighted_rkd_loss": 0.014404327888041735 + }, + { + "epoch": 2.5811965811965814, + "flow_matching_loss": 0.023364758491516112, + "grad_norm": 0.31836071610450745, + "learning_rate": 2.5093353128117996e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010980317369103432, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005202659475617111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02716329451650381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1510, + "weighted_flow_matching_loss": 0.023364758491516112, + "weighted_rkd_loss": 0.013581647258251905 + }, + { + "epoch": 2.5982905982905984, + "flow_matching_loss": 0.022494507301598788, + "grad_norm": 0.2116675227880478, + "learning_rate": 2.501974440473037e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010303075471892954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004565139859914779, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025171290896832944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1520, + "weighted_flow_matching_loss": 0.022494507301598788, + "weighted_rkd_loss": 0.012585645448416472 + }, + { + "epoch": 2.6153846153846154, + "flow_matching_loss": 0.026725856214761735, + "grad_norm": 0.24244584143161774, + "learning_rate": 2.4945697385263015e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011164265777915716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005043569020926952, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027372100763022898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1530, + "weighted_flow_matching_loss": 0.026725856214761735, + "weighted_rkd_loss": 0.013686050381511449 + }, + { + "epoch": 2.6324786324786325, + "flow_matching_loss": 0.02198776612058282, + "grad_norm": 0.22380423545837402, + "learning_rate": 2.487121530877243e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010480466578155757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004462654585950077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025423587858676912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1540, + "weighted_flow_matching_loss": 0.02198776612058282, + "weighted_rkd_loss": 0.012711793929338456 + }, + { + "epoch": 2.6495726495726495, + "flow_matching_loss": 0.02080122269690037, + "grad_norm": 0.24043220281600952, + "learning_rate": 2.4796301433345943e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008758699102327228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038610855117440224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02137848371639848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1550, + "weighted_flow_matching_loss": 0.02080122269690037, + "weighted_rkd_loss": 0.01068924185819924 + }, + { + "epoch": 2.6666666666666665, + "flow_matching_loss": 0.023161404207348822, + "grad_norm": 0.24361535906791687, + "learning_rate": 2.4720959035959148e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011636437475681305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005215252633206546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028488127514719963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1560, + "weighted_flow_matching_loss": 0.023161404207348822, + "weighted_rkd_loss": 0.014244063757359982 + }, + { + "epoch": 2.683760683760684, + "flow_matching_loss": 0.02135396469384432, + "grad_norm": 0.24413393437862396, + "learning_rate": 2.464519141233257e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010051483660936356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042491775238886476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024352144729346038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1570, + "weighted_flow_matching_loss": 0.02135396469384432, + "weighted_rkd_loss": 0.012176072364673019 + }, + { + "epoch": 2.700854700854701, + "flow_matching_loss": 0.022891308926045895, + "grad_norm": 0.2563834488391876, + "learning_rate": 2.456900187678754e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009299258328974247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004032260063104332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02263077683746815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1580, + "weighted_flow_matching_loss": 0.022891308926045895, + "weighted_rkd_loss": 0.011315388418734074 + }, + { + "epoch": 2.717948717948718, + "flow_matching_loss": 0.019574376195669173, + "grad_norm": 0.2142282873392105, + "learning_rate": 2.4492393762101147e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01088754371739924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005167186912149191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02694227434694767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1590, + "weighted_flow_matching_loss": 0.019574376195669173, + "weighted_rkd_loss": 0.013471137173473835 + }, + { + "epoch": 2.735042735042735, + "flow_matching_loss": 0.023829315975308418, + "grad_norm": 0.2677934169769287, + "learning_rate": 2.4415370419360508e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010912666656076908, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004734570253640413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026559903100132943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1600, + "weighted_flow_matching_loss": 0.023829315975308418, + "weighted_rkd_loss": 0.013279951550066471 + }, + { + "epoch": 2.752136752136752, + "flow_matching_loss": 0.022046511620283128, + "grad_norm": 0.18143510818481445, + "learning_rate": 2.4337935217816157e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009173155901953578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004238010593689978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022584322560578584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1610, + "weighted_flow_matching_loss": 0.022046511620283128, + "weighted_rkd_loss": 0.011292161280289292 + }, + { + "epoch": 2.769230769230769, + "flow_matching_loss": 0.0256638266146183, + "grad_norm": 0.26739150285720825, + "learning_rate": 2.4260091544734665e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009626912791281939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045542302541434765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02380805565044284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1620, + "weighted_flow_matching_loss": 0.0256638266146183, + "weighted_rkd_loss": 0.01190402782522142 + }, + { + "epoch": 2.786324786324786, + "flow_matching_loss": 0.02314116433262825, + "grad_norm": 0.27141332626342773, + "learning_rate": 2.4181842805250466e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00963348699733615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044304505223408345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02369742449373007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1630, + "weighted_flow_matching_loss": 0.02314116433262825, + "weighted_rkd_loss": 0.011848712246865034 + }, + { + "epoch": 2.8034188034188032, + "flow_matching_loss": 0.02181963212788105, + "grad_norm": 0.20872725546360016, + "learning_rate": 2.4103192422216923e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009032536670565606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004314871900714934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022379945125430824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1640, + "weighted_flow_matching_loss": 0.02181963212788105, + "weighted_rkd_loss": 0.011189972562715412 + }, + { + "epoch": 2.8205128205128203, + "flow_matching_loss": 0.022701138444244862, + "grad_norm": 0.1778896450996399, + "learning_rate": 2.402414383605658e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010498900059610605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045100216753780845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02550782198086381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1650, + "weighted_flow_matching_loss": 0.022701138444244862, + "weighted_rkd_loss": 0.012753910990431905 + }, + { + "epoch": 2.8376068376068377, + "flow_matching_loss": 0.024045399948954584, + "grad_norm": 0.25071728229522705, + "learning_rate": 2.3944700504610674e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00897721559740603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003891855734400451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021846286952495575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1660, + "weighted_flow_matching_loss": 0.024045399948954584, + "weighted_rkd_loss": 0.010923143476247787 + }, + { + "epoch": 2.8547008547008548, + "flow_matching_loss": 0.02234038049355149, + "grad_norm": 0.20470502972602844, + "learning_rate": 2.3864865902987883e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011603549402207137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051022290484979745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028309327736496925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1670, + "weighted_flow_matching_loss": 0.02234038049355149, + "weighted_rkd_loss": 0.014154663868248462 + }, + { + "epoch": 2.871794871794872, + "flow_matching_loss": 0.023738479986786843, + "grad_norm": 0.30578407645225525, + "learning_rate": 2.378464352341231e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008975940523669124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003991176700219512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02194305770099163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1680, + "weighted_flow_matching_loss": 0.023738479986786843, + "weighted_rkd_loss": 0.010971528850495816 + }, + { + "epoch": 2.888888888888889, + "flow_matching_loss": 0.022554483730345964, + "grad_norm": 0.1951245367527008, + "learning_rate": 2.3704036875070728e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009812219673767686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004091479280032217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023715918976813556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1690, + "weighted_flow_matching_loss": 0.022554483730345964, + "weighted_rkd_loss": 0.011857959488406778 + }, + { + "epoch": 2.905982905982906, + "flow_matching_loss": 0.02242042701691389, + "grad_norm": 0.2974187135696411, + "learning_rate": 2.362304948395906e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010304574249312281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004178128391504288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024787276796996594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1700, + "weighted_flow_matching_loss": 0.02242042701691389, + "weighted_rkd_loss": 0.012393638398498297 + }, + { + "epoch": 2.9230769230769234, + "flow_matching_loss": 0.023922652564942838, + "grad_norm": 0.18772614002227783, + "learning_rate": 2.3541684892728157e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009445326868444681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004097510641440749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022988164238631724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1710, + "weighted_flow_matching_loss": 0.023922652564942838, + "weighted_rkd_loss": 0.011494082119315862 + }, + { + "epoch": 2.9401709401709404, + "flow_matching_loss": 0.026682638563215733, + "grad_norm": 0.2330038994550705, + "learning_rate": 2.345994666052883e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009895666409283877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00440065455622971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024191987700760365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1720, + "weighted_flow_matching_loss": 0.026682638563215733, + "weighted_rkd_loss": 0.012095993850380182 + }, + { + "epoch": 2.9572649572649574, + "flow_matching_loss": 0.02124050781130791, + "grad_norm": 0.19767366349697113, + "learning_rate": 2.337783836285614e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008275231532752513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003853382426314056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020403845328837635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1730, + "weighted_flow_matching_loss": 0.02124050781130791, + "weighted_rkd_loss": 0.010201922664418817 + }, + { + "epoch": 2.9743589743589745, + "flow_matching_loss": 0.02081846250221133, + "grad_norm": 0.307229608297348, + "learning_rate": 2.3295363591393026e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009578318148851395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040630410658195615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02321967724710703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1740, + "weighted_flow_matching_loss": 0.02081846250221133, + "weighted_rkd_loss": 0.011609838623553515 + }, + { + "epoch": 2.9914529914529915, + "flow_matching_loss": 0.0220365384593606, + "grad_norm": 0.19888272881507874, + "learning_rate": 2.3212525953853174e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009073050040751696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003744899039156735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021890999004244804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1750, + "weighted_flow_matching_loss": 0.0220365384593606, + "weighted_rkd_loss": 0.010945499502122402 + }, + { + "epoch": 3.0085470085470085, + "flow_matching_loss": 0.020825703255832197, + "grad_norm": 0.1918894499540329, + "learning_rate": 2.3129329073823206e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009859509905800224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045070439809933305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024226063676178456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1760, + "weighted_flow_matching_loss": 0.020825703255832197, + "weighted_rkd_loss": 0.012113031838089228 + }, + { + "epoch": 3.0256410256410255, + "flow_matching_loss": 0.02404018212109804, + "grad_norm": 0.19095730781555176, + "learning_rate": 2.3045776590604175e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008907624846324324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004010485555045307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021825735550373794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1770, + "weighted_flow_matching_loss": 0.02404018212109804, + "weighted_rkd_loss": 0.010912867775186897 + }, + { + "epoch": 3.0427350427350426, + "flow_matching_loss": 0.02037848560139537, + "grad_norm": 0.23975929617881775, + "learning_rate": 2.296187215905237e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009382487926632165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042326563969254495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022997632250189783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1780, + "weighted_flow_matching_loss": 0.02037848560139537, + "weighted_rkd_loss": 0.011498816125094891 + }, + { + "epoch": 3.0598290598290596, + "flow_matching_loss": 0.02206662083044648, + "grad_norm": 0.14885295927524567, + "learning_rate": 2.2877619449419437e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008644994953647256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035399044398218395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020829894207417966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1790, + "weighted_flow_matching_loss": 0.02206662083044648, + "weighted_rkd_loss": 0.010414947103708983 + }, + { + "epoch": 3.076923076923077, + "flow_matching_loss": 0.020480931922793388, + "grad_norm": 0.20446527004241943, + "learning_rate": 2.2793022147191843e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009374905563890933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042403008788824085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02299011219292879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1800, + "weighted_flow_matching_loss": 0.020480931922793388, + "weighted_rkd_loss": 0.011495056096464396 + }, + { + "epoch": 3.094017094017094, + "flow_matching_loss": 0.021702952776104213, + "grad_norm": 0.19124248623847961, + "learning_rate": 2.2708083952929634e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009283293923363089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004115194361656904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022681782208383083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1810, + "weighted_flow_matching_loss": 0.021702952776104213, + "weighted_rkd_loss": 0.011340891104191541 + }, + { + "epoch": 3.111111111111111, + "flow_matching_loss": 0.023065047152340413, + "grad_norm": 0.16134604811668396, + "learning_rate": 2.2622808582104594e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008247178699821233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038086207583546638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020302978157997132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1820, + "weighted_flow_matching_loss": 0.023065047152340413, + "weighted_rkd_loss": 0.010151489078998566 + }, + { + "epoch": 3.128205128205128, + "flow_matching_loss": 0.020875094179064035, + "grad_norm": 0.21919263899326324, + "learning_rate": 2.2537199764937703e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00895846332423389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00419829033780843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022115217056125402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1830, + "weighted_flow_matching_loss": 0.020875094179064035, + "weighted_rkd_loss": 0.011057608528062701 + }, + { + "epoch": 3.1452991452991452, + "flow_matching_loss": 0.02138959923759103, + "grad_norm": 0.23678432404994965, + "learning_rate": 2.245126124623595e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007867638859897852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003307641646824777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019042919389903545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1840, + "weighted_flow_matching_loss": 0.02138959923759103, + "weighted_rkd_loss": 0.009521459694951773 + }, + { + "epoch": 3.1623931623931623, + "flow_matching_loss": 0.02178165577352047, + "grad_norm": 0.27445220947265625, + "learning_rate": 2.2364996785228547e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008633937686681747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004018479073420167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021286354307085274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1850, + "weighted_flow_matching_loss": 0.02178165577352047, + "weighted_rkd_loss": 0.010643177153542637 + }, + { + "epoch": 3.1794871794871793, + "flow_matching_loss": 0.021973855700343846, + "grad_norm": 0.2308157980442047, + "learning_rate": 2.227841015540248e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007728042174130678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032928862725384533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018748970329761507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1860, + "weighted_flow_matching_loss": 0.021973855700343846, + "weighted_rkd_loss": 0.009374485164880753 + }, + { + "epoch": 3.1965811965811968, + "flow_matching_loss": 0.021395375672727823, + "grad_norm": 0.21695679426193237, + "learning_rate": 2.2191505144337433e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077186715323477985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033147968002595007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018752139899879693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1870, + "weighted_flow_matching_loss": 0.021395375672727823, + "weighted_rkd_loss": 0.009376069949939846 + }, + { + "epoch": 3.213675213675214, + "flow_matching_loss": 0.021534157823771237, + "grad_norm": 0.1653919219970703, + "learning_rate": 2.210428555354013e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00908505842089653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035650128265842795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02173512978479266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1880, + "weighted_flow_matching_loss": 0.021534157823771237, + "weighted_rkd_loss": 0.01086756489239633 + }, + { + "epoch": 3.230769230769231, + "flow_matching_loss": 0.02106862384825945, + "grad_norm": 0.2326592653989792, + "learning_rate": 2.201675519827802e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009765352075919508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004243217711336911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023773921933025122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1890, + "weighted_flow_matching_loss": 0.02106862384825945, + "weighted_rkd_loss": 0.011886960966512561 + }, + { + "epoch": 3.247863247863248, + "flow_matching_loss": 0.021206633374094962, + "grad_norm": 0.19293636083602905, + "learning_rate": 2.1928917907412403e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007537059253081679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003404276235960424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018478394765406848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1900, + "weighted_flow_matching_loss": 0.021206633374094962, + "weighted_rkd_loss": 0.009239197382703424 + }, + { + "epoch": 3.264957264957265, + "flow_matching_loss": 0.02262264396995306, + "grad_norm": 0.20528379082679749, + "learning_rate": 2.1840777523230937e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008054467104375362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038436053320765494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019952539540827274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1910, + "weighted_flow_matching_loss": 0.02262264396995306, + "weighted_rkd_loss": 0.009976269770413637 + }, + { + "epoch": 3.282051282051282, + "flow_matching_loss": 0.022744439914822577, + "grad_norm": 0.18302470445632935, + "learning_rate": 2.1752337901279552e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011205578362569213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004998157313093543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027409313805401327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1920, + "weighted_flow_matching_loss": 0.022744439914822577, + "weighted_rkd_loss": 0.013704656902700663 + }, + { + "epoch": 3.299145299145299, + "flow_matching_loss": 0.021091637015342713, + "grad_norm": 0.27177563309669495, + "learning_rate": 2.166360291019383e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008100183354690671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032902485225349666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019490615092217924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1930, + "weighted_flow_matching_loss": 0.021091637015342713, + "weighted_rkd_loss": 0.009745307546108962 + }, + { + "epoch": 3.316239316239316, + "flow_matching_loss": 0.022644120268523693, + "grad_norm": 0.20682282745838165, + "learning_rate": 2.157457643152973e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009497429803013801, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00464288117364049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02363774077966809, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1940, + "weighted_flow_matching_loss": 0.022644120268523693, + "weighted_rkd_loss": 0.011818870389834046 + }, + { + "epoch": 3.3333333333333335, + "flow_matching_loss": 0.02207537516951561, + "grad_norm": 0.24180874228477478, + "learning_rate": 2.148526235959385e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008376868302002549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032720234477892516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020025760307908057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1950, + "weighted_flow_matching_loss": 0.02207537516951561, + "weighted_rkd_loss": 0.010012880153954028 + }, + { + "epoch": 3.3504273504273505, + "flow_matching_loss": 0.02177107408642769, + "grad_norm": 0.25179964303970337, + "learning_rate": 2.1395664601273016e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00934967128559947, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004402558598667383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023101901076734065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1960, + "weighted_flow_matching_loss": 0.02177107408642769, + "weighted_rkd_loss": 0.011550950538367032 + }, + { + "epoch": 3.3675213675213675, + "flow_matching_loss": 0.021467326395213603, + "grad_norm": 0.20472826063632965, + "learning_rate": 2.130578707586345e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00802918765693903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00377117128809914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01982954628765583, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1970, + "weighted_flow_matching_loss": 0.021467326395213603, + "weighted_rkd_loss": 0.009914773143827915 + }, + { + "epoch": 3.3846153846153846, + "flow_matching_loss": 0.022415131144225598, + "grad_norm": 0.18386898934841156, + "learning_rate": 2.1215633714899264e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008150833193212748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038708670064806936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020172533206641675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1980, + "weighted_flow_matching_loss": 0.022415131144225598, + "weighted_rkd_loss": 0.010086266603320838 + }, + { + "epoch": 3.4017094017094016, + "flow_matching_loss": 0.02110731778666377, + "grad_norm": 0.1889202892780304, + "learning_rate": 2.112520846198053e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00840694373473525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003788764402270317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020602651871740818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1990, + "weighted_flow_matching_loss": 0.02110731778666377, + "weighted_rkd_loss": 0.010301325935870409 + }, + { + "epoch": 3.4188034188034186, + "flow_matching_loss": 0.019866282120347023, + "grad_norm": 0.26760584115982056, + "learning_rate": 2.103451527260074e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011116993194445968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044356569647789005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02666964316740632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2000, + "weighted_flow_matching_loss": 0.019866282120347023, + "weighted_rkd_loss": 0.01333482158370316 + }, + { + "epoch": 3.435897435897436, + "flow_matching_loss": 0.018249011039733885, + "grad_norm": 0.23657004535198212, + "learning_rate": 2.094355811397381e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008862052159383894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038880287203937767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021612132992595436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2010, + "weighted_flow_matching_loss": 0.018249011039733885, + "weighted_rkd_loss": 0.010806066496297718 + }, + { + "epoch": 3.452991452991453, + "flow_matching_loss": 0.022592077776789666, + "grad_norm": 0.16656915843486786, + "learning_rate": 2.0852340964860522e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008196860598400236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035904660122469067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019984187465161084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2020, + "weighted_flow_matching_loss": 0.022592077776789666, + "weighted_rkd_loss": 0.009992093732580542 + }, + { + "epoch": 3.47008547008547, + "flow_matching_loss": 0.021013226732611655, + "grad_norm": 0.2137608677148819, + "learning_rate": 2.076086781539447e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008774524834007025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004004479385912418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021553528774529696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2030, + "weighted_flow_matching_loss": 0.021013226732611655, + "weighted_rkd_loss": 0.010776764387264848 + }, + { + "epoch": 3.4871794871794872, + "flow_matching_loss": 0.019182110857218503, + "grad_norm": 0.2665335237979889, + "learning_rate": 2.066914266690755e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008432052610442043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039183218264952305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020782427024096252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2040, + "weighted_flow_matching_loss": 0.019182110857218503, + "weighted_rkd_loss": 0.010391213512048126 + }, + { + "epoch": 3.5042735042735043, + "flow_matching_loss": 0.02038408424705267, + "grad_norm": 0.15456117689609528, + "learning_rate": 2.0577169531754906e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010282731615006924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004464774415828288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025030237529426812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2050, + "weighted_flow_matching_loss": 0.02038408424705267, + "weighted_rkd_loss": 0.012515118764713406 + }, + { + "epoch": 3.5213675213675213, + "flow_matching_loss": 0.021545465849339963, + "grad_norm": 0.1929696649312973, + "learning_rate": 2.048495243313942e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008824501093477011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036449615377932786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02129396367818117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2060, + "weighted_flow_matching_loss": 0.021545465849339963, + "weighted_rkd_loss": 0.010646981839090585 + }, + { + "epoch": 3.5384615384615383, + "flow_matching_loss": 0.018455321434885263, + "grad_norm": 0.23541535437107086, + "learning_rate": 2.039249540493572e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00999117074534297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003966285148635507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023948626406490803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2070, + "weighted_flow_matching_loss": 0.018455321434885263, + "weighted_rkd_loss": 0.011974313203245402 + }, + { + "epoch": 3.5555555555555554, + "flow_matching_loss": 0.018786207027733325, + "grad_norm": 0.22462044656276703, + "learning_rate": 2.0299802491513755e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00863970723003149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035249478416517375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02080436246469617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2080, + "weighted_flow_matching_loss": 0.018786207027733325, + "weighted_rkd_loss": 0.010402181232348085 + }, + { + "epoch": 3.5726495726495724, + "flow_matching_loss": 0.019480399787425995, + "grad_norm": 0.2128298580646515, + "learning_rate": 2.0206877747561828e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007677402580156922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003288325946778059, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01864313129335642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2090, + "weighted_flow_matching_loss": 0.019480399787425995, + "weighted_rkd_loss": 0.00932156564667821 + }, + { + "epoch": 3.58974358974359, + "flow_matching_loss": 0.01911817295476794, + "grad_norm": 0.23664461076259613, + "learning_rate": 2.011372523790927e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00922760833054781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004206936457194388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022662153095006944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2100, + "weighted_flow_matching_loss": 0.01911817295476794, + "weighted_rkd_loss": 0.011331076547503472 + }, + { + "epoch": 3.606837606837607, + "flow_matching_loss": 0.021250076405704022, + "grad_norm": 0.21973693370819092, + "learning_rate": 2.0020349037348643e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007971875090152025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003988684457726777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019932434521615507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2110, + "weighted_flow_matching_loss": 0.021250076405704022, + "weighted_rkd_loss": 0.009966217260807753 + }, + { + "epoch": 3.623931623931624, + "flow_matching_loss": 0.019656579475849867, + "grad_norm": 0.21634699404239655, + "learning_rate": 1.9926753230457443e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00955271115526557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003916142298839987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023021564818918705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2120, + "weighted_flow_matching_loss": 0.019656579475849867, + "weighted_rkd_loss": 0.011510782409459352 + }, + { + "epoch": 3.641025641025641, + "flow_matching_loss": 0.02276723589748144, + "grad_norm": 0.15986588597297668, + "learning_rate": 1.983294191141948e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075195180252194405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003291030717082322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01833006674423814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2130, + "weighted_flow_matching_loss": 0.02276723589748144, + "weighted_rkd_loss": 0.00916503337211907 + }, + { + "epoch": 3.658119658119658, + "flow_matching_loss": 0.020830532629042865, + "grad_norm": 0.21286843717098236, + "learning_rate": 1.973891918384575e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008494880562648178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003529363591223955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020519124809652567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2140, + "weighted_flow_matching_loss": 0.020830532629042865, + "weighted_rkd_loss": 0.010259562404826283 + }, + { + "epoch": 3.6752136752136755, + "flow_matching_loss": 0.023566334135830404, + "grad_norm": 0.24354811012744904, + "learning_rate": 1.9644689160594945e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008519923128187657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036325100576505066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02067235643044114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2150, + "weighted_flow_matching_loss": 0.023566334135830404, + "weighted_rkd_loss": 0.01033617821522057 + }, + { + "epoch": 3.6923076923076925, + "flow_matching_loss": 0.018900660425424577, + "grad_norm": 0.1920747309923172, + "learning_rate": 1.9550255963593553e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007615871122106909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003055795456748456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018287537898868323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2160, + "weighted_flow_matching_loss": 0.018900660425424577, + "weighted_rkd_loss": 0.009143768949434162 + }, + { + "epoch": 3.7094017094017095, + "flow_matching_loss": 0.01953559760004282, + "grad_norm": 0.19230477511882782, + "learning_rate": 1.9455623723655524e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067968220915645364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031586338998749854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016752277966588736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2170, + "weighted_flow_matching_loss": 0.01953559760004282, + "weighted_rkd_loss": 0.008376138983294368 + }, + { + "epoch": 3.7264957264957266, + "flow_matching_loss": 0.020689549297094344, + "grad_norm": 0.23367641866207123, + "learning_rate": 1.9360796580301606e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008445434179157019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003690579044632614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020581447705626487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2180, + "weighted_flow_matching_loss": 0.020689549297094344, + "weighted_rkd_loss": 0.010290723852813244 + }, + { + "epoch": 3.7435897435897436, + "flow_matching_loss": 0.020847134850919248, + "grad_norm": 0.2316247820854187, + "learning_rate": 1.9265778681578245e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008731900528073312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034508655546233056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020914666634052993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2190, + "weighted_flow_matching_loss": 0.020847134850919248, + "weighted_rkd_loss": 0.010457333317026496 + }, + { + "epoch": 3.7606837606837606, + "flow_matching_loss": 0.02032330334186554, + "grad_norm": 0.17385540902614594, + "learning_rate": 1.9170574183876143e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007317671971395612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032924892380833625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017927832994610073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2200, + "weighted_flow_matching_loss": 0.02032330334186554, + "weighted_rkd_loss": 0.008963916497305036 + }, + { + "epoch": 3.7777777777777777, + "flow_matching_loss": 0.016899863351136447, + "grad_norm": 0.26192307472229004, + "learning_rate": 1.9075187251748452e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007569550909101963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003170735482126474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018309837207198143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2210, + "weighted_flow_matching_loss": 0.016899863351136447, + "weighted_rkd_loss": 0.009154918603599072 + }, + { + "epoch": 3.7948717948717947, + "flow_matching_loss": 0.018195388186722995, + "grad_norm": 0.25618091225624084, + "learning_rate": 1.8979622057728596e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007942789187654852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032951420871540904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019180720765143634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2220, + "weighted_flow_matching_loss": 0.018195388186722995, + "weighted_rkd_loss": 0.009590360382571817 + }, + { + "epoch": 3.8119658119658117, + "flow_matching_loss": 0.02252973020076752, + "grad_norm": 0.1578928828239441, + "learning_rate": 1.8883882782147748e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074423910584300755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003065627347677946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017950409650802614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2230, + "weighted_flow_matching_loss": 0.02252973020076752, + "weighted_rkd_loss": 0.008975204825401307 + }, + { + "epoch": 3.8290598290598292, + "flow_matching_loss": 0.02154372911900282, + "grad_norm": 0.20821666717529297, + "learning_rate": 1.8787973612951967e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00786780584603548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032656229101121425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01900123441591859, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2240, + "weighted_flow_matching_loss": 0.02154372911900282, + "weighted_rkd_loss": 0.009500617207959295 + }, + { + "epoch": 3.8461538461538463, + "flow_matching_loss": 0.020391280949115752, + "grad_norm": 0.17726753652095795, + "learning_rate": 1.8691898745519028e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007469217339530587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003028665063902736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01796709969639778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2250, + "weighted_flow_matching_loss": 0.020391280949115752, + "weighted_rkd_loss": 0.00898354984819889 + }, + { + "epoch": 3.8632478632478633, + "flow_matching_loss": 0.02223206451162696, + "grad_norm": 0.21495142579078674, + "learning_rate": 1.8595662382474863e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007301053707487881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003117244434542954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017719351686537266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2260, + "weighted_flow_matching_loss": 0.02223206451162696, + "weighted_rkd_loss": 0.008859675843268633 + }, + { + "epoch": 3.8803418803418803, + "flow_matching_loss": 0.01902555041015148, + "grad_norm": 0.20775483548641205, + "learning_rate": 1.8499268733509766e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073735327925533055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003244478744454682, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017991544399410486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2270, + "weighted_flow_matching_loss": 0.01902555041015148, + "weighted_rkd_loss": 0.008995772199705243 + }, + { + "epoch": 3.8974358974358974, + "flow_matching_loss": 0.021454056072980166, + "grad_norm": 0.19884011149406433, + "learning_rate": 1.84027220151942e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007698776153847575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034163655480369927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018813917879015207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2280, + "weighted_flow_matching_loss": 0.021454056072980166, + "weighted_rkd_loss": 0.009406958939507604 + }, + { + "epoch": 3.9145299145299144, + "flow_matching_loss": 0.020456366799771787, + "grad_norm": 0.21196779608726501, + "learning_rate": 1.83060264507944e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007814684696495534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031296714674681424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018759041093289852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2290, + "weighted_flow_matching_loss": 0.020456366799771787, + "weighted_rkd_loss": 0.009379520546644926 + }, + { + "epoch": 3.931623931623932, + "flow_matching_loss": 0.019748034421354532, + "grad_norm": 0.21080656349658966, + "learning_rate": 1.82091862700876e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007885579485446215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030657083028927446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018836867436766623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2300, + "weighted_flow_matching_loss": 0.019748034421354532, + "weighted_rkd_loss": 0.009418433718383312 + }, + { + "epoch": 3.948717948717949, + "flow_matching_loss": 0.021026203967630865, + "grad_norm": 0.19430294632911682, + "learning_rate": 1.8112205709177023e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007923238072544336, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032659520395100115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019112428463995457, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2310, + "weighted_flow_matching_loss": 0.021026203967630865, + "weighted_rkd_loss": 0.009556214231997729 + }, + { + "epoch": 3.965811965811966, + "flow_matching_loss": 0.019944813381880522, + "grad_norm": 0.32070183753967285, + "learning_rate": 1.8015089010306585e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009109126357361674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003980800183489919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022199052944779397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2320, + "weighted_flow_matching_loss": 0.019944813381880522, + "weighted_rkd_loss": 0.011099526472389699 + }, + { + "epoch": 3.982905982905983, + "flow_matching_loss": 0.019468062836676837, + "grad_norm": 0.25614234805107117, + "learning_rate": 1.7917840421675298e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008891282556578516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003779213014058769, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021561777964234353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2330, + "weighted_flow_matching_loss": 0.019468062836676837, + "weighted_rkd_loss": 0.010780888982117176 + }, + { + "epoch": 4.0, + "flow_matching_loss": 0.017012250050902365, + "grad_norm": 0.16915491223335266, + "learning_rate": 1.7820464197251474e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006800329172983765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029427450615912678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01654340336099267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2340, + "weighted_flow_matching_loss": 0.017012250050902365, + "weighted_rkd_loss": 0.008271701680496336 + }, + { + "epoch": 4.017094017094017, + "flow_matching_loss": 0.023955489788204432, + "grad_norm": 0.22803570330142975, + "learning_rate": 1.7722964596586627e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006895763799548149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002875331451650709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016666858922690153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2350, + "weighted_flow_matching_loss": 0.023955489788204432, + "weighted_rkd_loss": 0.008333429461345077 + }, + { + "epoch": 4.034188034188034, + "flow_matching_loss": 0.017703853361308575, + "grad_norm": 0.19360429048538208, + "learning_rate": 1.7625345884629144e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007683675503358245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003063351823948324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018430702853947878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2360, + "weighted_flow_matching_loss": 0.017703853361308575, + "weighted_rkd_loss": 0.009215351426973939 + }, + { + "epoch": 4.051282051282051, + "flow_matching_loss": 0.02146550826728344, + "grad_norm": 0.15965017676353455, + "learning_rate": 1.7527612331537722e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066555246710777284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002755923569202423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016066972725093366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2370, + "weighted_flow_matching_loss": 0.02146550826728344, + "weighted_rkd_loss": 0.008033486362546683 + }, + { + "epoch": 4.068376068376068, + "flow_matching_loss": 0.01993013881146908, + "grad_norm": 0.20406599342823029, + "learning_rate": 1.7429768212494593e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006303849536925555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027177043608389794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015325403492897748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2380, + "weighted_flow_matching_loss": 0.01993013881146908, + "weighted_rkd_loss": 0.007662701746448874 + }, + { + "epoch": 4.085470085470085, + "flow_matching_loss": 0.01773031707853079, + "grad_norm": 0.19607993960380554, + "learning_rate": 1.7331817807518497e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007189767109230161, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032816751045174897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017661209404468536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2390, + "weighted_flow_matching_loss": 0.01773031707853079, + "weighted_rkd_loss": 0.008830604702234268 + }, + { + "epoch": 4.102564102564102, + "flow_matching_loss": 0.018448451347649098, + "grad_norm": 0.1936192661523819, + "learning_rate": 1.723376540127745e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006861878465861082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031465153908357022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016870272066444158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2400, + "weighted_flow_matching_loss": 0.018448451347649098, + "weighted_rkd_loss": 0.008435136033222079 + }, + { + "epoch": 4.119658119658119, + "flow_matching_loss": 0.01903025442734361, + "grad_norm": 0.28160524368286133, + "learning_rate": 1.7135615282901357e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007035943772643804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002800643094815314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016872530709952115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2410, + "weighted_flow_matching_loss": 0.01903025442734361, + "weighted_rkd_loss": 0.008436265354976057 + }, + { + "epoch": 4.136752136752137, + "flow_matching_loss": 0.02104166429489851, + "grad_norm": 0.15420423448085785, + "learning_rate": 1.7037371745794358e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008220721362158656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032787902280688288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019720232859253884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2420, + "weighted_flow_matching_loss": 0.02104166429489851, + "weighted_rkd_loss": 0.009860116429626942 + }, + { + "epoch": 4.153846153846154, + "flow_matching_loss": 0.019651689007878302, + "grad_norm": 0.15077415108680725, + "learning_rate": 1.6939039087447026e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075453647645190355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031900717760436235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018280801270157097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2430, + "weighted_flow_matching_loss": 0.019651689007878302, + "weighted_rkd_loss": 0.009140400635078549 + }, + { + "epoch": 4.170940170940171, + "flow_matching_loss": 0.020570564549416304, + "grad_norm": 0.15573781728744507, + "learning_rate": 1.6840621609248397e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007544708624482155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003139568492770195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018228985741734503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2440, + "weighted_flow_matching_loss": 0.020570564549416304, + "weighted_rkd_loss": 0.009114492870867252 + }, + { + "epoch": 4.188034188034188, + "flow_matching_loss": 0.019456946291029453, + "grad_norm": 0.20012281835079193, + "learning_rate": 1.6742123616297794e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006495161354541779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027588428696617485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01574916560202837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2450, + "weighted_flow_matching_loss": 0.019456946291029453, + "weighted_rkd_loss": 0.007874582801014185 + }, + { + "epoch": 4.205128205128205, + "flow_matching_loss": 0.019614775478839875, + "grad_norm": 0.1647314727306366, + "learning_rate": 1.664354941721651e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006208974169567227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002630527503788471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01504847602918744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2460, + "weighted_flow_matching_loss": 0.019614775478839875, + "weighted_rkd_loss": 0.00752423801459372 + }, + { + "epoch": 4.222222222222222, + "flow_matching_loss": 0.0180529972538352, + "grad_norm": 0.1961066573858261, + "learning_rate": 1.6544903323959355e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073069647885859014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002847653639037162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017461583390831946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2470, + "weighted_flow_matching_loss": 0.0180529972538352, + "weighted_rkd_loss": 0.008730791695415973 + }, + { + "epoch": 4.239316239316239, + "flow_matching_loss": 0.020445212721824646, + "grad_norm": 0.22484345734119415, + "learning_rate": 1.6446189651626023e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006985000707209111, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031812989618629217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017151300329715014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2480, + "weighted_flow_matching_loss": 0.020445212721824646, + "weighted_rkd_loss": 0.008575650164857507 + }, + { + "epoch": 4.256410256410256, + "flow_matching_loss": 0.019169483799487352, + "grad_norm": 0.2385309785604477, + "learning_rate": 1.6347412718272316e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006840714206919074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025678937672637403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016249322239309548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2490, + "weighted_flow_matching_loss": 0.019169483799487352, + "weighted_rkd_loss": 0.008124661119654774 + }, + { + "epoch": 4.273504273504273, + "flow_matching_loss": 0.017080207541584968, + "grad_norm": 0.17671377956867218, + "learning_rate": 1.624857684472131e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007922574994154274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002863072499167174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018708222638815643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2500, + "weighted_flow_matching_loss": 0.017080207541584968, + "weighted_rkd_loss": 0.009354111319407821 + }, + { + "epoch": 4.2905982905982905, + "flow_matching_loss": 0.019747746549546717, + "grad_norm": 0.24573016166687012, + "learning_rate": 1.6149686354374283e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007329531060531735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030151148559525607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017674177046865226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2510, + "weighted_flow_matching_loss": 0.019747746549546717, + "weighted_rkd_loss": 0.008837088523432613 + }, + { + "epoch": 4.3076923076923075, + "flow_matching_loss": 0.018726011179387568, + "grad_norm": 0.22373729944229126, + "learning_rate": 1.605074557302165e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006631332472898066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002862296684179455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016124961618334054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2520, + "weighted_flow_matching_loss": 0.018726011179387568, + "weighted_rkd_loss": 0.008062480809167027 + }, + { + "epoch": 4.3247863247863245, + "flow_matching_loss": 0.021509300265461207, + "grad_norm": 0.1717628836631775, + "learning_rate": 1.59517588286537e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006188838463276625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025393213145434856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01491699842736125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2530, + "weighted_flow_matching_loss": 0.021509300265461207, + "weighted_rkd_loss": 0.007458499213680625 + }, + { + "epoch": 4.3418803418803416, + "flow_matching_loss": 0.020329602248966693, + "grad_norm": 0.18526530265808105, + "learning_rate": 1.5852730451271324e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007229596329852939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002875377354212105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01733456999063492, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2540, + "weighted_flow_matching_loss": 0.020329602248966693, + "weighted_rkd_loss": 0.00866728499531746 + }, + { + "epoch": 4.358974358974359, + "flow_matching_loss": 0.0184131451882422, + "grad_norm": 0.18414761126041412, + "learning_rate": 1.5753664772696546e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006622652034275233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026653855573385956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01591068971902132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2550, + "weighted_flow_matching_loss": 0.0184131451882422, + "weighted_rkd_loss": 0.00795534485951066 + }, + { + "epoch": 4.3760683760683765, + "flow_matching_loss": 0.02079639583826065, + "grad_norm": 0.18560418486595154, + "learning_rate": 1.565456612638307e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007964245975017548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003640728979371488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019569220952689648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2560, + "weighted_flow_matching_loss": 0.02079639583826065, + "weighted_rkd_loss": 0.009784610476344824 + }, + { + "epoch": 4.3931623931623935, + "flow_matching_loss": 0.01935685221105814, + "grad_norm": 0.20377367734909058, + "learning_rate": 1.5555438847226732e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062549715396016834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029534086119383574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015463351644575596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2570, + "weighted_flow_matching_loss": 0.01935685221105814, + "weighted_rkd_loss": 0.007731675822287798 + }, + { + "epoch": 4.410256410256411, + "flow_matching_loss": 0.018714569322764874, + "grad_norm": 0.16832947731018066, + "learning_rate": 1.545628727137584e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006530326465144754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027387462090700866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015799399092793466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2580, + "weighted_flow_matching_loss": 0.018714569322764874, + "weighted_rkd_loss": 0.007899699546396733 + }, + { + "epoch": 4.427350427350428, + "flow_matching_loss": 0.019943490531295537, + "grad_norm": 0.17356619238853455, + "learning_rate": 1.535711573604153e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006290619866922498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028495809994637964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015430820826441049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2590, + "weighted_flow_matching_loss": 0.019943490531295537, + "weighted_rkd_loss": 0.0077154104132205244 + }, + { + "epoch": 4.444444444444445, + "flow_matching_loss": 0.017166621424257754, + "grad_norm": 0.1952560544013977, + "learning_rate": 1.5257928579308032e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006391754373908043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028482529683969914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015631761588156223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2600, + "weighted_flow_matching_loss": 0.017166621424257754, + "weighted_rkd_loss": 0.007815880794078111 + }, + { + "epoch": 4.461538461538462, + "flow_matching_loss": 0.017560683749616147, + "grad_norm": 0.1988353580236435, + "learning_rate": 1.5158730139942902e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007574155880138278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033369460492394865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018485257867723703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2610, + "weighted_flow_matching_loss": 0.017560683749616147, + "weighted_rkd_loss": 0.009242628933861852 + }, + { + "epoch": 4.478632478632479, + "flow_matching_loss": 0.019688459113240242, + "grad_norm": 0.19683468341827393, + "learning_rate": 1.5059524757207239e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063693718519061806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031602281494997444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015898971911519767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2620, + "weighted_flow_matching_loss": 0.019688459113240242, + "weighted_rkd_loss": 0.007949485955759884 + }, + { + "epoch": 4.495726495726496, + "flow_matching_loss": 0.01867778431624174, + "grad_norm": 0.2284543812274933, + "learning_rate": 1.4960316770665866e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006279498664662242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026411876315250993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015200185030698777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2630, + "weighted_flow_matching_loss": 0.01867778431624174, + "weighted_rkd_loss": 0.007600092515349388 + }, + { + "epoch": 4.512820512820513, + "flow_matching_loss": 0.01726768594235182, + "grad_norm": 0.1803433746099472, + "learning_rate": 1.4861110519997495e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005782140232622623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024866451509296893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014050925616174936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2640, + "weighted_flow_matching_loss": 0.01726768594235182, + "weighted_rkd_loss": 0.007025462808087468 + }, + { + "epoch": 4.52991452991453, + "flow_matching_loss": 0.018991167843341827, + "grad_norm": 0.24880416691303253, + "learning_rate": 1.4761910344804913e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008187596360221506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00336456069489941, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01973975347355008, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2650, + "weighted_flow_matching_loss": 0.018991167843341827, + "weighted_rkd_loss": 0.00986987673677504 + }, + { + "epoch": 4.547008547008547, + "flow_matching_loss": 0.020480296481400727, + "grad_norm": 0.15083114802837372, + "learning_rate": 1.466272058442516e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006214395631104708, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026107010897248985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015039492212235928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2660, + "weighted_flow_matching_loss": 0.020480296481400727, + "weighted_rkd_loss": 0.007519746106117964 + }, + { + "epoch": 4.564102564102564, + "flow_matching_loss": 0.016016822867095472, + "grad_norm": 0.24289827048778534, + "learning_rate": 1.4563545577739676e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007098955847322941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003096250956878066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017294162698090077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2670, + "weighted_flow_matching_loss": 0.016016822867095472, + "weighted_rkd_loss": 0.008647081349045039 + }, + { + "epoch": 4.581196581196581, + "flow_matching_loss": 0.016501933615654708, + "grad_norm": 0.23007218539714813, + "learning_rate": 1.4464389662984543e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073565777856856585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026881229132413864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017401278391480444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2680, + "weighted_flow_matching_loss": 0.016501933615654708, + "weighted_rkd_loss": 0.008700639195740222 + }, + { + "epoch": 4.598290598290598, + "flow_matching_loss": 0.016896218247711658, + "grad_norm": 0.2506307065486908, + "learning_rate": 1.4365257177560684e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00605305025819689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002633159328252077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014739259891211987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2690, + "weighted_flow_matching_loss": 0.016896218247711658, + "weighted_rkd_loss": 0.007369629945605994 + }, + { + "epoch": 4.615384615384615, + "flow_matching_loss": 0.019083459116518497, + "grad_norm": 0.1666194647550583, + "learning_rate": 1.426615245784416e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006203994457609952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027952198404818775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015203209035098553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2700, + "weighted_flow_matching_loss": 0.019083459116518497, + "weighted_rkd_loss": 0.0076016045175492765 + }, + { + "epoch": 4.632478632478632, + "flow_matching_loss": 0.017737826984375715, + "grad_norm": 0.20131826400756836, + "learning_rate": 1.4167079838996468e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005915038753300905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021665282547473907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013996605761349201, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2710, + "weighted_flow_matching_loss": 0.017737826984375715, + "weighted_rkd_loss": 0.006998302880674601 + }, + { + "epoch": 4.64957264957265, + "flow_matching_loss": 0.01762369042262435, + "grad_norm": 0.2097741961479187, + "learning_rate": 1.4068043654774897e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066073834663257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024939873139373956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0157087542116642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2720, + "weighted_flow_matching_loss": 0.01762369042262435, + "weighted_rkd_loss": 0.0078543771058321 + }, + { + "epoch": 4.666666666666667, + "flow_matching_loss": 0.017110300622880457, + "grad_norm": 0.1796332746744156, + "learning_rate": 1.3969048237342992e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006355493422597646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002593577420338988, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015304564125835895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2730, + "weighted_flow_matching_loss": 0.017110300622880457, + "weighted_rkd_loss": 0.007652282062917948 + }, + { + "epoch": 4.683760683760684, + "flow_matching_loss": 0.020454686228185893, + "grad_norm": 0.18637192249298096, + "learning_rate": 1.3870097917081011e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053040543105453255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002318077115342021, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012926185876131058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2740, + "weighted_flow_matching_loss": 0.020454686228185893, + "weighted_rkd_loss": 0.006463092938065529 + }, + { + "epoch": 4.700854700854701, + "flow_matching_loss": 0.016615460719913246, + "grad_norm": 0.18428291380405426, + "learning_rate": 1.3771197022396508e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006608629692345857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031284614000469447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016345720645040272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2750, + "weighted_flow_matching_loss": 0.016615460719913246, + "weighted_rkd_loss": 0.008172860322520136 + }, + { + "epoch": 4.717948717948718, + "flow_matching_loss": 0.018700057454407216, + "grad_norm": 0.15285761654376984, + "learning_rate": 1.3672349879535015e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008110217843204736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003631356067489833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019851791579276322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2760, + "weighted_flow_matching_loss": 0.018700057454407216, + "weighted_rkd_loss": 0.009925895789638161 + }, + { + "epoch": 4.735042735042735, + "flow_matching_loss": 0.016846880689263342, + "grad_norm": 0.18228141963481903, + "learning_rate": 1.3573560812390781e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005839766724966467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002764455007854849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01444398844614625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2770, + "weighted_flow_matching_loss": 0.016846880689263342, + "weighted_rkd_loss": 0.007221994223073125 + }, + { + "epoch": 4.752136752136752, + "flow_matching_loss": 0.017821814958006145, + "grad_norm": 0.17568399012088776, + "learning_rate": 1.3474834142317635e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006056840531527996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024255528813228013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014539233967661857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2780, + "weighted_flow_matching_loss": 0.017821814958006145, + "weighted_rkd_loss": 0.007269616983830929 + }, + { + "epoch": 4.769230769230769, + "flow_matching_loss": 0.01895545031875372, + "grad_norm": 0.12359953671693802, + "learning_rate": 1.3376174187939955e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006296288664452731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026273989700712264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015219976473599673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2790, + "weighted_flow_matching_loss": 0.01895545031875372, + "weighted_rkd_loss": 0.0076099882367998365 + }, + { + "epoch": 4.786324786324786, + "flow_matching_loss": 0.01891577858477831, + "grad_norm": 0.15576958656311035, + "learning_rate": 1.3277585264963761e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005916955787688494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002357321046292782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014191232435405255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2800, + "weighted_flow_matching_loss": 0.01891577858477831, + "weighted_rkd_loss": 0.0070956162177026275 + }, + { + "epoch": 4.803418803418803, + "flow_matching_loss": 0.018333879485726356, + "grad_norm": 0.22275856137275696, + "learning_rate": 1.3179071685987928e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006870065536350012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026655447203665973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016405675560235977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2810, + "weighted_flow_matching_loss": 0.018333879485726356, + "weighted_rkd_loss": 0.008202837780117989 + }, + { + "epoch": 4.82051282051282, + "flow_matching_loss": 0.019350035302340984, + "grad_norm": 0.18743370473384857, + "learning_rate": 1.308063776031554e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00618456348311156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021372218267060815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014506348874419928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2820, + "weighted_flow_matching_loss": 0.019350035302340984, + "weighted_rkd_loss": 0.007253174437209964 + }, + { + "epoch": 4.837606837606837, + "flow_matching_loss": 0.016245046723634006, + "grad_norm": 0.20033948123455048, + "learning_rate": 1.2982287793765389e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066329317167401316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023788953665643932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015644758567214014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2830, + "weighted_flow_matching_loss": 0.016245046723634006, + "weighted_rkd_loss": 0.007822379283607007 + }, + { + "epoch": 4.854700854700854, + "flow_matching_loss": 0.016420507896691562, + "grad_norm": 0.1553906500339508, + "learning_rate": 1.2884026088483616e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058491195552051066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002730893774423748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014429132873192429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2840, + "weighted_flow_matching_loss": 0.016420507896691562, + "weighted_rkd_loss": 0.007214566436596214 + }, + { + "epoch": 4.871794871794872, + "flow_matching_loss": 0.021777248568832876, + "grad_norm": 0.17237120866775513, + "learning_rate": 1.2785856942755543e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005849478789605201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024362468044273554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014135204302147032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2850, + "weighted_flow_matching_loss": 0.021777248568832876, + "weighted_rkd_loss": 0.007067602151073516 + }, + { + "epoch": 4.888888888888889, + "flow_matching_loss": 0.018680552765727044, + "grad_norm": 0.19036276638507843, + "learning_rate": 1.2687784650817616e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00545023640152067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002043917903210968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012944390717893838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2860, + "weighted_flow_matching_loss": 0.018680552765727044, + "weighted_rkd_loss": 0.006472195358946919 + }, + { + "epoch": 4.905982905982906, + "flow_matching_loss": 0.0194564713165164, + "grad_norm": 0.18401499092578888, + "learning_rate": 1.2589813502669603e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007179317437112332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028180433670058846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01717667831107974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2870, + "weighted_flow_matching_loss": 0.0194564713165164, + "weighted_rkd_loss": 0.00858833915553987 + }, + { + "epoch": 4.923076923076923, + "flow_matching_loss": 0.018249529600143432, + "grad_norm": 0.15625852346420288, + "learning_rate": 1.2491947783886907e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005985748581588268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025901356828399004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014561632834374904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2880, + "weighted_flow_matching_loss": 0.018249529600143432, + "weighted_rkd_loss": 0.007280816417187452 + }, + { + "epoch": 4.94017094017094, + "flow_matching_loss": 0.019586964324116708, + "grad_norm": 0.19170287251472473, + "learning_rate": 1.2394191775433106e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006269854912534356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026000387500971557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015139748714864254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2890, + "weighted_flow_matching_loss": 0.019586964324116708, + "weighted_rkd_loss": 0.007569874357432127 + }, + { + "epoch": 4.957264957264957, + "flow_matching_loss": 0.018550141993910075, + "grad_norm": 0.217694953083992, + "learning_rate": 1.2296549753472696e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063949782401323315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023791401297785343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015169096365571021, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2900, + "weighted_flow_matching_loss": 0.018550141993910075, + "weighted_rkd_loss": 0.007584548182785511 + }, + { + "epoch": 4.9743589743589745, + "flow_matching_loss": 0.01646477123722434, + "grad_norm": 0.1519484668970108, + "learning_rate": 1.2199025989184032e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006408982444554567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00265109499450773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015469059906899929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2910, + "weighted_flow_matching_loss": 0.01646477123722434, + "weighted_rkd_loss": 0.007734529953449964 + }, + { + "epoch": 4.9914529914529915, + "flow_matching_loss": 0.01712658563628793, + "grad_norm": 0.17936837673187256, + "learning_rate": 1.21016247485725e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005534190079197287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002666365006007254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01373474532738328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2920, + "weighted_flow_matching_loss": 0.01712658563628793, + "weighted_rkd_loss": 0.00686737266369164 + }, + { + "epoch": 5.0085470085470085, + "flow_matching_loss": 0.017433699034154416, + "grad_norm": 0.19425033032894135, + "learning_rate": 1.2004350292283897e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005776865407824516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002266028500162065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013819759339094162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2930, + "weighted_flow_matching_loss": 0.017433699034154416, + "weighted_rkd_loss": 0.006909879669547081 + }, + { + "epoch": 5.0256410256410255, + "flow_matching_loss": 0.01623713644221425, + "grad_norm": 0.15402722358703613, + "learning_rate": 1.1907206875418068e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005933634308166802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025461736135184766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01441344222985208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2940, + "weighted_flow_matching_loss": 0.01623713644221425, + "weighted_rkd_loss": 0.00720672111492604 + }, + { + "epoch": 5.042735042735043, + "flow_matching_loss": 0.01722178068011999, + "grad_norm": 0.17161601781845093, + "learning_rate": 1.1810198747342769e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006133737275376916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025711448513902726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014838619343936443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2950, + "weighted_flow_matching_loss": 0.01722178068011999, + "weighted_rkd_loss": 0.007419309671968221 + }, + { + "epoch": 5.05982905982906, + "flow_matching_loss": 0.01563998917117715, + "grad_norm": 0.17408081889152527, + "learning_rate": 1.1713330151507786e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006273146439343691, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002796279580798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01534257223829627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2960, + "weighted_flow_matching_loss": 0.01563998917117715, + "weighted_rkd_loss": 0.007671286119148135 + }, + { + "epoch": 5.076923076923077, + "flow_matching_loss": 0.018053919449448584, + "grad_norm": 0.18406715989112854, + "learning_rate": 1.161660532525931e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006768700829707086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002817504294216633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01635490609332919, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2970, + "weighted_flow_matching_loss": 0.018053919449448584, + "weighted_rkd_loss": 0.008177453046664595 + }, + { + "epoch": 5.094017094017094, + "flow_matching_loss": 0.015317833703011274, + "grad_norm": 0.1968030482530594, + "learning_rate": 1.1520028499654584e-05, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00502970430534333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002209481818135828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01226889044046402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2980, + "weighted_flow_matching_loss": 0.015317833703011274, + "weighted_rkd_loss": 0.00613444522023201 + }, + { + "epoch": 5.111111111111111, + "flow_matching_loss": 0.016654190700501205, + "grad_norm": 0.1656157225370407, + "learning_rate": 1.1423603899276833e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005547607317566872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024255677242763342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013520782347768545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2990, + "weighted_flow_matching_loss": 0.016654190700501205, + "weighted_rkd_loss": 0.006760391173884273 + }, + { + "epoch": 5.128205128205128, + "flow_matching_loss": 0.018000909499824046, + "grad_norm": 0.1773122400045395, + "learning_rate": 1.1327335742050442e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006210618140175939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002675327961333096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01509656421840191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3000, + "weighted_flow_matching_loss": 0.018000909499824046, + "weighted_rkd_loss": 0.007548282109200955 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/config.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4740d0455bef31c469ee5dab6d4a47d64df61665 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.5, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": null, + "rkd_loss_weight_start": 0.1, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..23dcfa871a88e56280f945357aef0296bd071711 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.577927505176127, + -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-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a3758b2dac9cab1274647a0a3e4647bb197ec7fc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6f702e66e76c9c2a5773af5ec6b03a2aabb911ab03f04ce0b6ef1d2a8dca66a +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/model.safetensors.index.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/optimizer.pt b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/optimizer.pt new file mode 100644 index 0000000000000000000000000000000000000000..3292d41d6de52d96457953383c3a6ecd95a7a718 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/optimizer.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:71ea010ed0f29364f1c68857d57567ab605a6f67f9ed1c20d250dfe8df4665e2 +size 1624159435 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/trainer_state.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/checkpoint-5000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..0cd5d9dddfed31febe3a6266273e129b46b096c9 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_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": 0.20608119964599608, + "grad_norm": 1.4477208852767944, + "learning_rate": 1.08e-06, + "loss": 0.4085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1441098630428314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11659819930791855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.4048179268836975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10, + "weighted_flow_matching_loss": 0.20608119964599608, + "weighted_rkd_loss": 0.20240896344184875 + }, + { + "epoch": 0.03418803418803419, + "flow_matching_loss": 0.22261241227388381, + "grad_norm": 1.1392409801483154, + "learning_rate": 2.28e-06, + "loss": 0.4131, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.13591156229376794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10920900553464889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.38103213608264924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20, + "weighted_flow_matching_loss": 0.22261241227388381, + "weighted_rkd_loss": 0.19051606804132462 + }, + { + "epoch": 0.05128205128205128, + "flow_matching_loss": 0.20390308946371077, + "grad_norm": 0.9336247444152832, + "learning_rate": 3.48e-06, + "loss": 0.388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1328342378139496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1025419719517231, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3682104468345642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 30, + "weighted_flow_matching_loss": 0.20390308946371077, + "weighted_rkd_loss": 0.1841052234172821 + }, + { + "epoch": 0.06837606837606838, + "flow_matching_loss": 0.2105909526348114, + "grad_norm": 0.8749905824661255, + "learning_rate": 4.68e-06, + "loss": 0.3894, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12913469299674035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09934469908475876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.35761408507823944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 40, + "weighted_flow_matching_loss": 0.2105909526348114, + "weighted_rkd_loss": 0.17880704253911972 + }, + { + "epoch": 0.08547008547008547, + "flow_matching_loss": 0.1939498320221901, + "grad_norm": 0.8234826326370239, + "learning_rate": 5.8800000000000005e-06, + "loss": 0.3703, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1269106835126877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09883174449205398, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3526531159877777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 50, + "weighted_flow_matching_loss": 0.1939498320221901, + "weighted_rkd_loss": 0.17632655799388885 + }, + { + "epoch": 0.10256410256410256, + "flow_matching_loss": 0.15859663859009743, + "grad_norm": 0.7916533946990967, + "learning_rate": 7.08e-06, + "loss": 0.3244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.12063853591680526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09041309282183647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3316901683807373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 60, + "weighted_flow_matching_loss": 0.15859663859009743, + "weighted_rkd_loss": 0.16584508419036864 + }, + { + "epoch": 0.11965811965811966, + "flow_matching_loss": 0.15783419758081435, + "grad_norm": 0.8823524117469788, + "learning_rate": 8.28e-06, + "loss": 0.3089, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11185461878776551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0784363217651844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30214556455612185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 70, + "weighted_flow_matching_loss": 0.15783419758081435, + "weighted_rkd_loss": 0.15107278227806092 + }, + { + "epoch": 0.13675213675213677, + "flow_matching_loss": 0.13894206359982492, + "grad_norm": 0.941081702709198, + "learning_rate": 9.48e-06, + "loss": 0.2816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1060655303299427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0731036301702261, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.28523469567298887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 80, + "weighted_flow_matching_loss": 0.13894206359982492, + "weighted_rkd_loss": 0.14261734783649443 + }, + { + "epoch": 0.15384615384615385, + "flow_matching_loss": 0.12293280363082885, + "grad_norm": 0.8303114771842957, + "learning_rate": 1.068e-05, + "loss": 0.258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10080260336399079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06849895156919956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.270104156434536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 90, + "weighted_flow_matching_loss": 0.12293280363082885, + "weighted_rkd_loss": 0.135052078217268 + }, + { + "epoch": 0.17094017094017094, + "flow_matching_loss": 0.1046357847750187, + "grad_norm": 0.7039380669593811, + "learning_rate": 1.1880000000000001e-05, + "loss": 0.2292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09367859661579132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06181581430137158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24917300790548325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 100, + "weighted_flow_matching_loss": 0.1046357847750187, + "weighted_rkd_loss": 0.12458650395274162 + }, + { + "epoch": 0.18803418803418803, + "flow_matching_loss": 0.09870888739824295, + "grad_norm": 0.6019426584243774, + "learning_rate": 1.308e-05, + "loss": 0.2142, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08723139390349388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05651620887219906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23097899854183196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 110, + "weighted_flow_matching_loss": 0.09870888739824295, + "weighted_rkd_loss": 0.11548949927091598 + }, + { + "epoch": 0.20512820512820512, + "flow_matching_loss": 0.08241323307156563, + "grad_norm": 0.6689212918281555, + "learning_rate": 1.428e-05, + "loss": 0.197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08734692260622978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0545398335903883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2292336791753769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 120, + "weighted_flow_matching_loss": 0.08241323307156563, + "weighted_rkd_loss": 0.11461683958768845 + }, + { + "epoch": 0.2222222222222222, + "flow_matching_loss": 0.0766731396317482, + "grad_norm": 0.578585684299469, + "learning_rate": 1.548e-05, + "loss": 0.1806, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07932209149003029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.049220692738890647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20786487609148024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 130, + "weighted_flow_matching_loss": 0.0766731396317482, + "weighted_rkd_loss": 0.10393243804574012 + }, + { + "epoch": 0.23931623931623933, + "flow_matching_loss": 0.08649998903274536, + "grad_norm": 0.6135573983192444, + "learning_rate": 1.6680000000000003e-05, + "loss": 0.1817, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07297333218157291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0444162305444479, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1903628960251808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 140, + "weighted_flow_matching_loss": 0.08649998903274536, + "weighted_rkd_loss": 0.0951814480125904 + }, + { + "epoch": 0.2564102564102564, + "flow_matching_loss": 0.07530920952558517, + "grad_norm": 0.6404624581336975, + "learning_rate": 1.7879999999999998e-05, + "loss": 0.1619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06675466485321521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.039724042639136314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17323337346315384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 150, + "weighted_flow_matching_loss": 0.07530920952558517, + "weighted_rkd_loss": 0.08661668673157692 + }, + { + "epoch": 0.27350427350427353, + "flow_matching_loss": 0.07482291832566261, + "grad_norm": 0.614500880241394, + "learning_rate": 1.908e-05, + "loss": 0.1624, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0673632200807333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.040419120155274865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17514555901288986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 160, + "weighted_flow_matching_loss": 0.07482291832566261, + "weighted_rkd_loss": 0.08757277950644493 + }, + { + "epoch": 0.2905982905982906, + "flow_matching_loss": 0.07167641781270503, + "grad_norm": 0.5213647484779358, + "learning_rate": 2.0280000000000002e-05, + "loss": 0.1554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0648277897387743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03784646335989237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16750204414129258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 170, + "weighted_flow_matching_loss": 0.07167641781270503, + "weighted_rkd_loss": 0.08375102207064629 + }, + { + "epoch": 0.3076923076923077, + "flow_matching_loss": 0.06723086908459663, + "grad_norm": 0.5122416019439697, + "learning_rate": 2.148e-05, + "loss": 0.144, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05934748239815235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034743427857756616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15343839228153228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 180, + "weighted_flow_matching_loss": 0.06723086908459663, + "weighted_rkd_loss": 0.07671919614076614 + }, + { + "epoch": 0.3247863247863248, + "flow_matching_loss": 0.06515696458518505, + "grad_norm": 0.5083960890769958, + "learning_rate": 2.268e-05, + "loss": 0.1374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.055695648118853566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033016927726566794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14440822303295137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 190, + "weighted_flow_matching_loss": 0.06515696458518505, + "weighted_rkd_loss": 0.07220411151647568 + }, + { + "epoch": 0.3418803418803419, + "flow_matching_loss": 0.06426175646483898, + "grad_norm": 0.5611753463745117, + "learning_rate": 2.3880000000000002e-05, + "loss": 0.1338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.053841910883784296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03130413573235273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13898795917630197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 200, + "weighted_flow_matching_loss": 0.06426175646483898, + "weighted_rkd_loss": 0.06949397958815098 + }, + { + "epoch": 0.358974358974359, + "flow_matching_loss": 0.061613532900810244, + "grad_norm": 0.4719548523426056, + "learning_rate": 2.508e-05, + "loss": 0.1246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.048743893206119535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02855243906378746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1260402247309685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 210, + "weighted_flow_matching_loss": 0.061613532900810244, + "weighted_rkd_loss": 0.06302011236548424 + }, + { + "epoch": 0.37606837606837606, + "flow_matching_loss": 0.058244922384619716, + "grad_norm": 0.489758163690567, + "learning_rate": 2.628e-05, + "loss": 0.1197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04813622012734413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026654975302517415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12292741760611534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 220, + "weighted_flow_matching_loss": 0.058244922384619716, + "weighted_rkd_loss": 0.06146370880305767 + }, + { + "epoch": 0.39316239316239315, + "flow_matching_loss": 0.05867415480315685, + "grad_norm": 0.5043800473213196, + "learning_rate": 2.748e-05, + "loss": 0.1202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04817534200847149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02667076177895069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1230214461684227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 230, + "weighted_flow_matching_loss": 0.05867415480315685, + "weighted_rkd_loss": 0.06151072308421135 + }, + { + "epoch": 0.41025641025641024, + "flow_matching_loss": 0.059197704493999484, + "grad_norm": 0.4292409420013428, + "learning_rate": 2.868e-05, + "loss": 0.1177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04579513482749462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025458005629479884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11704827472567558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 240, + "weighted_flow_matching_loss": 0.059197704493999484, + "weighted_rkd_loss": 0.05852413736283779 + }, + { + "epoch": 0.42735042735042733, + "flow_matching_loss": 0.05531383752822876, + "grad_norm": 0.43816468119621277, + "learning_rate": 2.9880000000000002e-05, + "loss": 0.1148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.046195313706994055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026640130765736104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1190307579934597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 250, + "weighted_flow_matching_loss": 0.05531383752822876, + "weighted_rkd_loss": 0.05951537899672985 + }, + { + "epoch": 0.4444444444444444, + "flow_matching_loss": 0.05735364258289337, + "grad_norm": 0.4437594413757324, + "learning_rate": 2.9999734259635676e-05, + "loss": 0.1043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03679828383028507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020236621610820293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09383318945765495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 260, + "weighted_flow_matching_loss": 0.05735364258289337, + "weighted_rkd_loss": 0.046916594728827477 + }, + { + "epoch": 0.46153846153846156, + "flow_matching_loss": 0.059609442576766015, + "grad_norm": 0.5054996013641357, + "learning_rate": 2.9998815663057245e-05, + "loss": 0.1117, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041301059536635876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021624139230698346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10422625914216041, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 270, + "weighted_flow_matching_loss": 0.059609442576766015, + "weighted_rkd_loss": 0.052113129571080205 + }, + { + "epoch": 0.47863247863247865, + "flow_matching_loss": 0.05288516990840435, + "grad_norm": 0.48650580644607544, + "learning_rate": 2.999724096969228e-05, + "loss": 0.1047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04097450710833073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021585627086460592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10353464037179946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 280, + "weighted_flow_matching_loss": 0.05288516990840435, + "weighted_rkd_loss": 0.05176732018589973 + }, + { + "epoch": 0.49572649572649574, + "flow_matching_loss": 0.0536977294832468, + "grad_norm": 0.42119359970092773, + "learning_rate": 2.9995010248422984e-05, + "loss": 0.1018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.038251850940287116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01963985590264201, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09614355638623237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 290, + "weighted_flow_matching_loss": 0.0536977294832468, + "weighted_rkd_loss": 0.048071778193116185 + }, + { + "epoch": 0.5128205128205128, + "flow_matching_loss": 0.048235632851719855, + "grad_norm": 0.4279158115386963, + "learning_rate": 2.999212359682832e-05, + "loss": 0.1011, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04124627523124218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023284840025007726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10577739253640175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 300, + "weighted_flow_matching_loss": 0.048235632851719855, + "weighted_rkd_loss": 0.05288869626820088 + }, + { + "epoch": 0.5299145299145299, + "flow_matching_loss": 0.04915821589529514, + "grad_norm": 0.4583510756492615, + "learning_rate": 2.9988581141179787e-05, + "loss": 0.0909, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03272451851516962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018078654445707797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08352769166231155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 310, + "weighted_flow_matching_loss": 0.04915821589529514, + "weighted_rkd_loss": 0.04176384583115578 + }, + { + "epoch": 0.5470085470085471, + "flow_matching_loss": 0.0478827927261591, + "grad_norm": 0.4557395875453949, + "learning_rate": 2.998438303643587e-05, + "loss": 0.0943, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03672756850719452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01931789694353938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09277303442358971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 320, + "weighted_flow_matching_loss": 0.0478827927261591, + "weighted_rkd_loss": 0.046386517211794855 + }, + { + "epoch": 0.5641025641025641, + "flow_matching_loss": 0.048558805510401724, + "grad_norm": 0.4305928349494934, + "learning_rate": 2.997952946623527e-05, + "loss": 0.0936, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03552434463053942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019004741962999106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09005343131721019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 330, + "weighted_flow_matching_loss": 0.048558805510401724, + "weighted_rkd_loss": 0.045026715658605096 + }, + { + "epoch": 0.5811965811965812, + "flow_matching_loss": 0.05098775625228882, + "grad_norm": 0.40920335054397583, + "learning_rate": 2.9974020642888875e-05, + "loss": 0.0918, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03249767124652862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01662399461492896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08161933720111847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 340, + "weighted_flow_matching_loss": 0.05098775625228882, + "weighted_rkd_loss": 0.040809668600559235 + }, + { + "epoch": 0.5982905982905983, + "flow_matching_loss": 0.047236426174640654, + "grad_norm": 0.4020651578903198, + "learning_rate": 2.9967856807370458e-05, + "loss": 0.0876, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032064920477569106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016578875295817853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08070871606469154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 350, + "weighted_flow_matching_loss": 0.047236426174640654, + "weighted_rkd_loss": 0.04035435803234577 + }, + { + "epoch": 0.6153846153846154, + "flow_matching_loss": 0.049100768938660624, + "grad_norm": 0.410824179649353, + "learning_rate": 2.996103822930615e-05, + "loss": 0.0887, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03171112034469843, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015830404963344334, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07925264574587346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 360, + "weighted_flow_matching_loss": 0.049100768938660624, + "weighted_rkd_loss": 0.03962632287293673 + }, + { + "epoch": 0.6324786324786325, + "flow_matching_loss": 0.04570407420396805, + "grad_norm": 0.41649773716926575, + "learning_rate": 2.9953565206962653e-05, + "loss": 0.0868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03249792978167534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017130458448082208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08212631791830063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 370, + "weighted_flow_matching_loss": 0.04570407420396805, + "weighted_rkd_loss": 0.04106315895915032 + }, + { + "epoch": 0.6495726495726496, + "flow_matching_loss": 0.04507583752274513, + "grad_norm": 0.41268017888069153, + "learning_rate": 2.9945438067234172e-05, + "loss": 0.0817, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029481365904212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014191387593746186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07315411977469921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 380, + "weighted_flow_matching_loss": 0.04507583752274513, + "weighted_rkd_loss": 0.036577059887349606 + }, + { + "epoch": 0.6666666666666666, + "flow_matching_loss": 0.04255103617906571, + "grad_norm": 0.38128671050071716, + "learning_rate": 2.9936657165628124e-05, + "loss": 0.0831, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032352514937520026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016295951325446367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08100098073482513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 390, + "weighted_flow_matching_loss": 0.04255103617906571, + "weighted_rkd_loss": 0.040500490367412566 + }, + { + "epoch": 0.6837606837606838, + "flow_matching_loss": 0.04447304848581553, + "grad_norm": 0.36115458607673645, + "learning_rate": 2.992722288624959e-05, + "loss": 0.0812, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02881138101220131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01579931229352951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07342207506299019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 400, + "weighted_flow_matching_loss": 0.04447304848581553, + "weighted_rkd_loss": 0.03671103753149509 + }, + { + "epoch": 0.7008547008547008, + "flow_matching_loss": 0.04228375982493162, + "grad_norm": 0.27871495485305786, + "learning_rate": 2.9917135641784515e-05, + "loss": 0.0782, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028570003621280194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014606839325278998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07174684777855873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 410, + "weighted_flow_matching_loss": 0.04228375982493162, + "weighted_rkd_loss": 0.03587342388927937 + }, + { + "epoch": 0.717948717948718, + "flow_matching_loss": 0.044394811242818834, + "grad_norm": 0.43788713216781616, + "learning_rate": 2.9906395873481655e-05, + "loss": 0.0769, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0254867784678936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013939346186816692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06491290293633938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 420, + "weighted_flow_matching_loss": 0.044394811242818834, + "weighted_rkd_loss": 0.03245645146816969 + }, + { + "epoch": 0.7350427350427351, + "flow_matching_loss": 0.04304563030600548, + "grad_norm": 0.32673171162605286, + "learning_rate": 2.989500405113326e-05, + "loss": 0.0757, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026353750936686994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012691938411444425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06539944037795067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 430, + "weighted_flow_matching_loss": 0.04304563030600548, + "weighted_rkd_loss": 0.03269972018897534 + }, + { + "epoch": 0.7521367521367521, + "flow_matching_loss": 0.04123639427125454, + "grad_norm": 0.33855384588241577, + "learning_rate": 2.9882960673054547e-05, + "loss": 0.0736, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02597996760159731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012860519997775554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06482045613229274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 440, + "weighted_flow_matching_loss": 0.04123639427125454, + "weighted_rkd_loss": 0.03241022806614637 + }, + { + "epoch": 0.7692307692307693, + "flow_matching_loss": 0.04132717214524746, + "grad_norm": 0.3718223571777344, + "learning_rate": 2.9870266266061886e-05, + "loss": 0.0739, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026512700691819192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012095902115106583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06512130312621593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 450, + "weighted_flow_matching_loss": 0.04132717214524746, + "weighted_rkd_loss": 0.032560651563107966 + }, + { + "epoch": 0.7863247863247863, + "flow_matching_loss": 0.042681030184030536, + "grad_norm": 0.3572709560394287, + "learning_rate": 2.9856921385449766e-05, + "loss": 0.0774, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02780984025448561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013740665093064309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06936034522950649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 460, + "weighted_flow_matching_loss": 0.042681030184030536, + "weighted_rkd_loss": 0.034680172614753244 + }, + { + "epoch": 0.8034188034188035, + "flow_matching_loss": 0.042522648349404335, + "grad_norm": 0.3548585772514343, + "learning_rate": 2.9842926614966502e-05, + "loss": 0.0715, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023204243928194045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011549634858965874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057958122715353964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 470, + "weighted_flow_matching_loss": 0.042522648349404335, + "weighted_rkd_loss": 0.028979061357676982 + }, + { + "epoch": 0.8205128205128205, + "flow_matching_loss": 0.03533885534852743, + "grad_norm": 0.3358217775821686, + "learning_rate": 2.982828256678868e-05, + "loss": 0.0665, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024509145691990853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01339802285656333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06241631470620632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 480, + "weighted_flow_matching_loss": 0.03533885534852743, + "weighted_rkd_loss": 0.03120815735310316 + }, + { + "epoch": 0.8376068376068376, + "flow_matching_loss": 0.0407237846404314, + "grad_norm": 0.40146124362945557, + "learning_rate": 2.9812989881494412e-05, + "loss": 0.0751, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027627827040851115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013586688507348298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06884234212338924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 490, + "weighted_flow_matching_loss": 0.0407237846404314, + "weighted_rkd_loss": 0.03442117106169462 + }, + { + "epoch": 0.8547008547008547, + "flow_matching_loss": 0.042021133750677106, + "grad_norm": 0.2975722849369049, + "learning_rate": 2.9797049228035296e-05, + "loss": 0.0724, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02429118975996971, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01214608708396554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060728466510772704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 500, + "weighted_flow_matching_loss": 0.042021133750677106, + "weighted_rkd_loss": 0.030364233255386352 + }, + { + "epoch": 0.8717948717948718, + "flow_matching_loss": 0.03959287479519844, + "grad_norm": 0.37908297777175903, + "learning_rate": 2.978046130370715e-05, + "loss": 0.0687, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02314175870269537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011852209083735942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0581357266753912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 510, + "weighted_flow_matching_loss": 0.03959287479519844, + "weighted_rkd_loss": 0.0290678633376956 + }, + { + "epoch": 0.8888888888888888, + "flow_matching_loss": 0.038183963485062124, + "grad_norm": 0.35484832525253296, + "learning_rate": 2.9763226834119535e-05, + "loss": 0.0666, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022654735110700132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01152317225933075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05683264210820198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 520, + "weighted_flow_matching_loss": 0.038183963485062124, + "weighted_rkd_loss": 0.02841632105410099 + }, + { + "epoch": 0.905982905982906, + "flow_matching_loss": 0.03838554341346025, + "grad_norm": 0.3829350471496582, + "learning_rate": 2.9745346573163963e-05, + "loss": 0.0678, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02331932317465544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012097928486764432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.058736575022339824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 530, + "weighted_flow_matching_loss": 0.03838554341346025, + "weighted_rkd_loss": 0.029368287511169912 + }, + { + "epoch": 0.9230769230769231, + "flow_matching_loss": 0.0373695008456707, + "grad_norm": 0.39486923813819885, + "learning_rate": 2.9726821302980987e-05, + "loss": 0.0647, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02153742704540491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011604637745767832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054679491370916364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 540, + "weighted_flow_matching_loss": 0.0373695008456707, + "weighted_rkd_loss": 0.027339745685458182 + }, + { + "epoch": 0.9401709401709402, + "flow_matching_loss": 0.03420814126729965, + "grad_norm": 0.30264589190483093, + "learning_rate": 2.9707651833925935e-05, + "loss": 0.0593, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020150486007332803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009803551714867353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050104524567723276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 550, + "weighted_flow_matching_loss": 0.03420814126729965, + "weighted_rkd_loss": 0.025052262283861638 + }, + { + "epoch": 0.9572649572649573, + "flow_matching_loss": 0.03828359097242355, + "grad_norm": 0.27153193950653076, + "learning_rate": 2.968783900453348e-05, + "loss": 0.0632, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02007452305406332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009622522396966815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049771569296717645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 560, + "weighted_flow_matching_loss": 0.03828359097242355, + "weighted_rkd_loss": 0.024885784648358823 + }, + { + "epoch": 0.9743589743589743, + "flow_matching_loss": 0.03746143244206905, + "grad_norm": 0.28983309864997864, + "learning_rate": 2.9667383681480968e-05, + "loss": 0.0648, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02210656674578786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010438397945836186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05465153120458126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 570, + "weighted_flow_matching_loss": 0.03746143244206905, + "weighted_rkd_loss": 0.02732576560229063 + }, + { + "epoch": 0.9914529914529915, + "flow_matching_loss": 0.035928660072386265, + "grad_norm": 0.30345749855041504, + "learning_rate": 2.9646286759550493e-05, + "loss": 0.0628, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021595945302397012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010564278997480869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05375616922974587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 580, + "weighted_flow_matching_loss": 0.035928660072386265, + "weighted_rkd_loss": 0.026878084614872934 + }, + { + "epoch": 1.0085470085470085, + "flow_matching_loss": 0.03552918881177902, + "grad_norm": 0.24723964929580688, + "learning_rate": 2.9624549161589764e-05, + "loss": 0.0581, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018401246704161166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008261607168242336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0450641006231308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 590, + "weighted_flow_matching_loss": 0.03552918881177902, + "weighted_rkd_loss": 0.0225320503115654 + }, + { + "epoch": 1.0256410256410255, + "flow_matching_loss": 0.03407776262611151, + "grad_norm": 0.30071568489074707, + "learning_rate": 2.9602171838471745e-05, + "loss": 0.0579, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019089525938034056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009438153170049191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04761720485985279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 600, + "weighted_flow_matching_loss": 0.03407776262611151, + "weighted_rkd_loss": 0.023808602429926395 + }, + { + "epoch": 1.0427350427350428, + "flow_matching_loss": 0.03366336897015572, + "grad_norm": 0.302155464887619, + "learning_rate": 2.957915576905303e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02271198071539402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011890822742134332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057314783707261084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 610, + "weighted_flow_matching_loss": 0.03366336897015572, + "weighted_rkd_loss": 0.028657391853630542 + }, + { + "epoch": 1.0598290598290598, + "flow_matching_loss": 0.03334424737840891, + "grad_norm": 0.31046783924102783, + "learning_rate": 2.9555501960131058e-05, + "loss": 0.0577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019524640496820213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009662961680442094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04871224202215672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 620, + "weighted_flow_matching_loss": 0.03334424737840891, + "weighted_rkd_loss": 0.02435612101107836 + }, + { + "epoch": 1.0769230769230769, + "flow_matching_loss": 0.03312153853476048, + "grad_norm": 0.28929588198661804, + "learning_rate": 2.9531211446400068e-05, + "loss": 0.0552, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017922693397849798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00835406300611794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044199449568986894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 630, + "weighted_flow_matching_loss": 0.03312153853476048, + "weighted_rkd_loss": 0.022099724784493447 + }, + { + "epoch": 1.0940170940170941, + "flow_matching_loss": 0.031021049618721007, + "grad_norm": 0.29626306891441345, + "learning_rate": 2.9506285290405813e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018444733694195747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008001519087702035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0448909867554903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 640, + "weighted_flow_matching_loss": 0.031021049618721007, + "weighted_rkd_loss": 0.02244549337774515 + }, + { + "epoch": 1.1111111111111112, + "flow_matching_loss": 0.032624239847064015, + "grad_norm": 0.2606738209724426, + "learning_rate": 2.948072458249911e-05, + "loss": 0.0572, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01954005965963006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010045465733855962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04912558533251286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 650, + "weighted_flow_matching_loss": 0.032624239847064015, + "weighted_rkd_loss": 0.02456279266625643 + }, + { + "epoch": 1.1282051282051282, + "flow_matching_loss": 0.0357012577354908, + "grad_norm": 0.35832640528678894, + "learning_rate": 2.9454530440788125e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01891506966203451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008951532281935215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04678167216479778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 660, + "weighted_flow_matching_loss": 0.0357012577354908, + "weighted_rkd_loss": 0.02339083608239889 + }, + { + "epoch": 1.1452991452991452, + "flow_matching_loss": 0.035934435203671455, + "grad_norm": 0.3307911455631256, + "learning_rate": 2.9427704011089474e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018776980880647897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008833648543804884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046387610211968425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 670, + "weighted_flow_matching_loss": 0.035934435203671455, + "weighted_rkd_loss": 0.023193805105984212 + }, + { + "epoch": 1.1623931623931625, + "flow_matching_loss": 0.035431265085935595, + "grad_norm": 0.32725879549980164, + "learning_rate": 2.94002464668781e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018233825359493493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008717586379498243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045185237750411035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 680, + "weighted_flow_matching_loss": 0.035431265085935595, + "weighted_rkd_loss": 0.022592618875205518 + }, + { + "epoch": 1.1794871794871795, + "flow_matching_loss": 0.03349699769169092, + "grad_norm": 0.3060956299304962, + "learning_rate": 2.9372159009235936e-05, + "loss": 0.0549, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017201956920325757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008407466765493154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042811380326747896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 690, + "weighted_flow_matching_loss": 0.03349699769169092, + "weighted_rkd_loss": 0.021405690163373948 + }, + { + "epoch": 1.1965811965811965, + "flow_matching_loss": 0.037000064924359324, + "grad_norm": 0.2392890304327011, + "learning_rate": 2.9343442866799365e-05, + "loss": 0.0589, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01774967508390546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008203081460669637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04370243176817894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 700, + "weighted_flow_matching_loss": 0.037000064924359324, + "weighted_rkd_loss": 0.02185121588408947 + }, + { + "epoch": 1.2136752136752136, + "flow_matching_loss": 0.0321138696745038, + "grad_norm": 0.2744523286819458, + "learning_rate": 2.9314099295705495e-05, + "loss": 0.0545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01787513755261898, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008945518126711249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04469579309225082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 710, + "weighted_flow_matching_loss": 0.0321138696745038, + "weighted_rkd_loss": 0.02234789654612541 + }, + { + "epoch": 1.2307692307692308, + "flow_matching_loss": 0.03291467633098364, + "grad_norm": 0.27376607060432434, + "learning_rate": 2.928412957953718e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0181625971570611, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008992884447798133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04531807880848646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 720, + "weighted_flow_matching_loss": 0.03291467633098364, + "weighted_rkd_loss": 0.02265903940424323 + }, + { + "epoch": 1.2478632478632479, + "flow_matching_loss": 0.031597768515348436, + "grad_norm": 0.29815053939819336, + "learning_rate": 2.925353502926689e-05, + "loss": 0.0513, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015860845521092414, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007615697477012873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03933738861232996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 730, + "weighted_flow_matching_loss": 0.031597768515348436, + "weighted_rkd_loss": 0.01966869430616498 + }, + { + "epoch": 1.264957264957265, + "flow_matching_loss": 0.03209205660969019, + "grad_norm": 0.29948389530181885, + "learning_rate": 2.922231698319937e-05, + "loss": 0.0538, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017416194640100004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008499196963384748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04333158619701862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 740, + "weighted_flow_matching_loss": 0.03209205660969019, + "weighted_rkd_loss": 0.02166579309850931 + }, + { + "epoch": 1.282051282051282, + "flow_matching_loss": 0.02969713993370533, + "grad_norm": 0.2717636227607727, + "learning_rate": 2.9190476806913088e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0162487150169909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007376636797562242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039874066784977916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 750, + "weighted_flow_matching_loss": 0.02969713993370533, + "weighted_rkd_loss": 0.019937033392488958 + }, + { + "epoch": 1.2991452991452992, + "flow_matching_loss": 0.031202525086700916, + "grad_norm": 0.28742673993110657, + "learning_rate": 2.9158015893200504e-05, + "loss": 0.0521, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017281018663197756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007276421086862684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04183845836669207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 760, + "weighted_flow_matching_loss": 0.031202525086700916, + "weighted_rkd_loss": 0.020919229183346034 + }, + { + "epoch": 1.3162393162393162, + "flow_matching_loss": 0.03131227232515812, + "grad_norm": 0.28647977113723755, + "learning_rate": 2.9124935662007137e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01633641654625535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007625138806179166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040297971293330195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 770, + "weighted_flow_matching_loss": 0.03131227232515812, + "weighted_rkd_loss": 0.020148985646665098 + }, + { + "epoch": 1.3333333333333333, + "flow_matching_loss": 0.031031945906579494, + "grad_norm": 0.3045690655708313, + "learning_rate": 2.9091237560369467e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01676619527861476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007660567946732044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041192957758903505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 780, + "weighted_flow_matching_loss": 0.031031945906579494, + "weighted_rkd_loss": 0.020596478879451752 + }, + { + "epoch": 1.3504273504273505, + "flow_matching_loss": 0.03271263670176268, + "grad_norm": 0.2736513912677765, + "learning_rate": 2.905692306235163e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01463465103879571, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00730584766715765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03657514955848455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 790, + "weighted_flow_matching_loss": 0.03271263670176268, + "weighted_rkd_loss": 0.018287574779242276 + }, + { + "epoch": 1.3675213675213675, + "flow_matching_loss": 0.030691403523087503, + "grad_norm": 0.28391724824905396, + "learning_rate": 2.9021993668980925e-05, + "loss": 0.0507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016008828580379487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008079428132623434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040097085386514665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 800, + "weighted_flow_matching_loss": 0.030691403523087503, + "weighted_rkd_loss": 0.020048542693257333 + }, + { + "epoch": 1.3846153846153846, + "flow_matching_loss": 0.034423727355897424, + "grad_norm": 0.27780401706695557, + "learning_rate": 2.8986450908182185e-05, + "loss": 0.0546, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016429053619503975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007567548984661698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04042565673589706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 810, + "weighted_flow_matching_loss": 0.034423727355897424, + "weighted_rkd_loss": 0.02021282836794853 + }, + { + "epoch": 1.4017094017094016, + "flow_matching_loss": 0.03095700442790985, + "grad_norm": 0.22609981894493103, + "learning_rate": 2.8950296334710902e-05, + "loss": 0.0519, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01725160414353013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007418723125010729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041921931132674214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 820, + "weighted_flow_matching_loss": 0.03095700442790985, + "weighted_rkd_loss": 0.020960965566337107 + }, + { + "epoch": 1.4188034188034189, + "flow_matching_loss": 0.03410516288131475, + "grad_norm": 0.2484390139579773, + "learning_rate": 2.8913531530085248e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015002989396452903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007663191389292478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037669170275330545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 830, + "weighted_flow_matching_loss": 0.03410516288131475, + "weighted_rkd_loss": 0.018834585137665272 + }, + { + "epoch": 1.435897435897436, + "flow_matching_loss": 0.030957960151135923, + "grad_norm": 0.2720315754413605, + "learning_rate": 2.887615810251687e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016805268358439207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007622529938817024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04123306684195995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 840, + "weighted_flow_matching_loss": 0.030957960151135923, + "weighted_rkd_loss": 0.020616533420979976 + }, + { + "epoch": 1.452991452991453, + "flow_matching_loss": 0.030326111428439616, + "grad_norm": 0.2526715099811554, + "learning_rate": 2.8838177686840577e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014052478969097138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006443384615704417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03454834185540676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 850, + "weighted_flow_matching_loss": 0.030326111428439616, + "weighted_rkd_loss": 0.01727417092770338 + }, + { + "epoch": 1.4700854700854702, + "flow_matching_loss": 0.03168223239481449, + "grad_norm": 0.3726175129413605, + "learning_rate": 2.8799591944442776e-05, + "loss": 0.0518, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01616350244730711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007998800138011575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04032580554485321, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 860, + "weighted_flow_matching_loss": 0.03168223239481449, + "weighted_rkd_loss": 0.020162902772426605 + }, + { + "epoch": 1.4871794871794872, + "flow_matching_loss": 0.030286968126893042, + "grad_norm": 0.24847525358200073, + "learning_rate": 2.8760402563188848e-05, + "loss": 0.0486, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014799616392701864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007003285316750407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03660251796245575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 870, + "weighted_flow_matching_loss": 0.030286968126893042, + "weighted_rkd_loss": 0.018301258981227874 + }, + { + "epoch": 1.5042735042735043, + "flow_matching_loss": 0.029968923516571523, + "grad_norm": 0.31482070684432983, + "learning_rate": 2.8720611257349285e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01352565586566925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006156855309382081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03320816699415445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 880, + "weighted_flow_matching_loss": 0.029968923516571523, + "weighted_rkd_loss": 0.016604083497077225 + }, + { + "epoch": 1.5213675213675213, + "flow_matching_loss": 0.03208826966583729, + "grad_norm": 0.26682037115097046, + "learning_rate": 2.868021976752471e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012783072888851166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006290466827340424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031856612861156465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 890, + "weighted_flow_matching_loss": 0.03208826966583729, + "weighted_rkd_loss": 0.015928306430578233 + }, + { + "epoch": 1.5384615384615383, + "flow_matching_loss": 0.029574750922620296, + "grad_norm": 0.2852403223514557, + "learning_rate": 2.8639229860569738e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01526008890941739, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071847605984658005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03770493865013123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 900, + "weighted_flow_matching_loss": 0.029574750922620296, + "weighted_rkd_loss": 0.018852469325065613 + }, + { + "epoch": 1.5555555555555556, + "flow_matching_loss": 0.028749799728393553, + "grad_norm": 0.24580393731594086, + "learning_rate": 2.8597643329515688e-05, + "loss": 0.0447, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012961071636527776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006020331801846623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031942475400865075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 910, + "weighted_flow_matching_loss": 0.028749799728393553, + "weighted_rkd_loss": 0.015971237700432538 + }, + { + "epoch": 1.5726495726495726, + "flow_matching_loss": 0.032204396836459634, + "grad_norm": 0.20449334383010864, + "learning_rate": 2.855546199349214e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014906360022723674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006444963859394193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0362576836720109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 920, + "weighted_flow_matching_loss": 0.032204396836459634, + "weighted_rkd_loss": 0.01812884183600545 + }, + { + "epoch": 1.5897435897435899, + "flow_matching_loss": 0.029012378305196762, + "grad_norm": 0.2885199785232544, + "learning_rate": 2.85126876976474e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015990046598017216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007635087100788951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03961518015712499, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 930, + "weighted_flow_matching_loss": 0.029012378305196762, + "weighted_rkd_loss": 0.019807590078562497 + }, + { + "epoch": 1.606837606837607, + "flow_matching_loss": 0.0270179970189929, + "grad_norm": 0.2731475830078125, + "learning_rate": 2.8469322313067726e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014710118900984526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006910082930698991, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03633032031357288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 940, + "weighted_flow_matching_loss": 0.0270179970189929, + "weighted_rkd_loss": 0.01816516015678644 + }, + { + "epoch": 1.623931623931624, + "flow_matching_loss": 0.03037530407309532, + "grad_norm": 0.19273969531059265, + "learning_rate": 2.8425367736695522e-05, + "loss": 0.0482, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014435206167399883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00685276158619672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03572317436337471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 950, + "weighted_flow_matching_loss": 0.03037530407309532, + "weighted_rkd_loss": 0.017861587181687354 + }, + { + "epoch": 1.641025641025641, + "flow_matching_loss": 0.027571728639304637, + "grad_norm": 0.25559166073799133, + "learning_rate": 2.838082589124636e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012530108634382486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006172595894895494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031232813000679018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 960, + "weighted_flow_matching_loss": 0.027571728639304637, + "weighted_rkd_loss": 0.015616406500339509 + }, + { + "epoch": 1.658119658119658, + "flow_matching_loss": 0.026423130929470063, + "grad_norm": 0.2502352297306061, + "learning_rate": 2.8335698725124843e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013723570108413696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00633350177668035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0337806424126029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 970, + "weighted_flow_matching_loss": 0.026423130929470063, + "weighted_rkd_loss": 0.01689032120630145 + }, + { + "epoch": 1.6752136752136753, + "flow_matching_loss": 0.026968703232705594, + "grad_norm": 0.29313844442367554, + "learning_rate": 2.8289988212339422e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015268613491207362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006759832962416112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0372970599681139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 980, + "weighted_flow_matching_loss": 0.026968703232705594, + "weighted_rkd_loss": 0.01864852998405695 + }, + { + "epoch": 1.6923076923076923, + "flow_matching_loss": 0.025319616310298443, + "grad_norm": 0.2522859573364258, + "learning_rate": 2.8243696352415997e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014617102593183518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006836280273273587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03607048541307449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 990, + "weighted_flow_matching_loss": 0.025319616310298443, + "weighted_rkd_loss": 0.018035242706537245 + }, + { + "epoch": 1.7094017094017095, + "flow_matching_loss": 0.030002601444721222, + "grad_norm": 0.24149076640605927, + "learning_rate": 2.8196825170310495e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01380759608000517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006553543242625892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03416873514652252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1000, + "weighted_flow_matching_loss": 0.030002601444721222, + "weighted_rkd_loss": 0.01708436757326126 + }, + { + "epoch": 1.7264957264957266, + "flow_matching_loss": 0.02807784080505371, + "grad_norm": 0.25452151894569397, + "learning_rate": 2.8149376716320256e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014782947115600109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00671710423193872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03628299850970507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1010, + "weighted_flow_matching_loss": 0.02807784080505371, + "weighted_rkd_loss": 0.018141499254852533 + }, + { + "epoch": 1.7435897435897436, + "flow_matching_loss": 0.02710084058344364, + "grad_norm": 0.33143165707588196, + "learning_rate": 2.8101353065994373e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014535702299326658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006557283597066999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03562868814915419, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1020, + "weighted_flow_matching_loss": 0.02710084058344364, + "weighted_rkd_loss": 0.017814344074577094 + }, + { + "epoch": 1.7606837606837606, + "flow_matching_loss": 0.02593045849353075, + "grad_norm": 0.3127861022949219, + "learning_rate": 2.8052756320042885e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01566762672737241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007534478232264519, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03886973131448031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1030, + "weighted_flow_matching_loss": 0.02593045849353075, + "weighted_rkd_loss": 0.019434865657240154 + }, + { + "epoch": 1.7777777777777777, + "flow_matching_loss": 0.0273212568834424, + "grad_norm": 0.2180718183517456, + "learning_rate": 2.8003588604244893e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01422199010848999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00723591735586524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035679897479712966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1040, + "weighted_flow_matching_loss": 0.0273212568834424, + "weighted_rkd_loss": 0.017839948739856483 + }, + { + "epoch": 1.7948717948717947, + "flow_matching_loss": 0.027072438225150108, + "grad_norm": 0.2212599217891693, + "learning_rate": 2.7953852069355565e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013318956829607487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0059395078103989364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03257742114365101, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1050, + "weighted_flow_matching_loss": 0.027072438225150108, + "weighted_rkd_loss": 0.016288710571825505 + }, + { + "epoch": 1.811965811965812, + "flow_matching_loss": 0.026676180213689803, + "grad_norm": 0.2384062260389328, + "learning_rate": 2.790354889101206e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011322647333145142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005065834778361022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027711129188537596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1060, + "weighted_flow_matching_loss": 0.026676180213689803, + "weighted_rkd_loss": 0.013855564594268798 + }, + { + "epoch": 1.8290598290598292, + "flow_matching_loss": 0.028350574523210527, + "grad_norm": 0.26893094182014465, + "learning_rate": 2.7852681269638366e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01224717302247882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005544273066334426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030038618855178355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1070, + "weighted_flow_matching_loss": 0.028350574523210527, + "weighted_rkd_loss": 0.015019309427589177 + }, + { + "epoch": 1.8461538461538463, + "flow_matching_loss": 0.023936538957059383, + "grad_norm": 0.32209914922714233, + "learning_rate": 2.780125143034902e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012277705408632756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00516571004409343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02972112111747265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1080, + "weighted_flow_matching_loss": 0.023936538957059383, + "weighted_rkd_loss": 0.014860560558736325 + }, + { + "epoch": 1.8632478632478633, + "flow_matching_loss": 0.02863725144416094, + "grad_norm": 0.22413775324821472, + "learning_rate": 2.7749261622851812e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012101378757506609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005476271221414209, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029679028689861296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1090, + "weighted_flow_matching_loss": 0.02863725144416094, + "weighted_rkd_loss": 0.014839514344930648 + }, + { + "epoch": 1.8803418803418803, + "flow_matching_loss": 0.027668011374771595, + "grad_norm": 0.26382219791412354, + "learning_rate": 2.7696714121349335e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013915220554918051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005969888484105468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03380032908171415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1100, + "weighted_flow_matching_loss": 0.027668011374771595, + "weighted_rkd_loss": 0.016900164540857077 + }, + { + "epoch": 1.8974358974358974, + "flow_matching_loss": 0.02955530472099781, + "grad_norm": 0.27939197421073914, + "learning_rate": 2.764361122443954e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01350008575245738, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00708389279898256, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034084064327180384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1110, + "weighted_flow_matching_loss": 0.02955530472099781, + "weighted_rkd_loss": 0.017042032163590192 + }, + { + "epoch": 1.9145299145299144, + "flow_matching_loss": 0.02437609266489744, + "grad_norm": 0.2628627419471741, + "learning_rate": 2.758995525501517e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01297243358567357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005478202877566218, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03142307046800852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1120, + "weighted_flow_matching_loss": 0.02437609266489744, + "weighted_rkd_loss": 0.01571153523400426 + }, + { + "epoch": 1.9316239316239316, + "flow_matching_loss": 0.028911850042641164, + "grad_norm": 0.23939956724643707, + "learning_rate": 2.753574856016215e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012397017190232873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005886134272441268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03068016842007637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1130, + "weighted_flow_matching_loss": 0.028911850042641164, + "weighted_rkd_loss": 0.015340084210038185 + }, + { + "epoch": 1.9487179487179487, + "flow_matching_loss": 0.02508879117667675, + "grad_norm": 0.20714369416236877, + "learning_rate": 2.7480993511056913e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01294971164315939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063501600408926604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03224958349019289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1140, + "weighted_flow_matching_loss": 0.02508879117667675, + "weighted_rkd_loss": 0.016124791745096446 + }, + { + "epoch": 1.965811965811966, + "flow_matching_loss": 0.024527727998793124, + "grad_norm": 0.19642457365989685, + "learning_rate": 2.7425692502862684e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011007771594449878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052364618517458435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027252004854381083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1150, + "weighted_flow_matching_loss": 0.024527727998793124, + "weighted_rkd_loss": 0.013626002427190542 + }, + { + "epoch": 1.982905982905983, + "flow_matching_loss": 0.02737492471933365, + "grad_norm": 0.23660895228385925, + "learning_rate": 2.7369847954624728e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0125096146017313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061733081936836244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031192537397146225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1160, + "weighted_flow_matching_loss": 0.02737492471933365, + "weighted_rkd_loss": 0.015596268698573112 + }, + { + "epoch": 2.0, + "flow_matching_loss": 0.026675242371857165, + "grad_norm": 0.25652265548706055, + "learning_rate": 2.731346230916448e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013060001470148564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006334699317812919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03245470225811005, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1170, + "weighted_flow_matching_loss": 0.026675242371857165, + "weighted_rkd_loss": 0.016227351129055025 + }, + { + "epoch": 2.017094017094017, + "flow_matching_loss": 0.025632075406610964, + "grad_norm": 0.22836749255657196, + "learning_rate": 2.7256538032972746e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012878391984850168, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005053568026050925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030810352042317392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1180, + "weighted_flow_matching_loss": 0.025632075406610964, + "weighted_rkd_loss": 0.015405176021158696 + }, + { + "epoch": 2.034188034188034, + "flow_matching_loss": 0.027616259641945362, + "grad_norm": 0.22967053949832916, + "learning_rate": 2.7199077616101773e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012935011554509402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005700286431238055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03157030940055847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1190, + "weighted_flow_matching_loss": 0.027616259641945362, + "weighted_rkd_loss": 0.015785154700279237 + }, + { + "epoch": 2.051282051282051, + "flow_matching_loss": 0.02580515146255493, + "grad_norm": 0.2185813933610916, + "learning_rate": 2.714108357205634e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013044429570436477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005809265840798617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03189812451601028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1200, + "weighted_flow_matching_loss": 0.02580515146255493, + "weighted_rkd_loss": 0.01594906225800514 + }, + { + "epoch": 2.0683760683760686, + "flow_matching_loss": 0.02779768891632557, + "grad_norm": 0.21726644039154053, + "learning_rate": 2.7082558437683803e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012998757604509591, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005653172452002764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03165068775415421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1210, + "weighted_flow_matching_loss": 0.02779768891632557, + "weighted_rkd_loss": 0.015825343877077104 + }, + { + "epoch": 2.0854700854700856, + "flow_matching_loss": 0.0263330465182662, + "grad_norm": 0.22308462858200073, + "learning_rate": 2.702350477306315e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011554498877376317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005302437697537243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028411435894668102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1220, + "weighted_flow_matching_loss": 0.0263330465182662, + "weighted_rkd_loss": 0.014205717947334051 + }, + { + "epoch": 2.1025641025641026, + "flow_matching_loss": 0.02595714498311281, + "grad_norm": 0.22663040459156036, + "learning_rate": 2.6963925161392963e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011306988494470715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004992392286658287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027606369741261006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1230, + "weighted_flow_matching_loss": 0.02595714498311281, + "weighted_rkd_loss": 0.013803184870630503 + }, + { + "epoch": 2.1196581196581197, + "flow_matching_loss": 0.024864679016172887, + "grad_norm": 0.3029778301715851, + "learning_rate": 2.6903822208878476e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01236183405853808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00573265366256237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030456322245299816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1240, + "weighted_flow_matching_loss": 0.024864679016172887, + "weighted_rkd_loss": 0.015228161122649908 + }, + { + "epoch": 2.1367521367521367, + "flow_matching_loss": 0.02325292509049177, + "grad_norm": 0.2267565280199051, + "learning_rate": 2.6843198544617535e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011847892682999373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00517009780742228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028865883499383925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1250, + "weighted_flow_matching_loss": 0.02325292509049177, + "weighted_rkd_loss": 0.014432941749691963 + }, + { + "epoch": 2.1538461538461537, + "flow_matching_loss": 0.02737160213291645, + "grad_norm": 0.20818153023719788, + "learning_rate": 2.6782056820485612e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012656896701082587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00568916043266654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031002953462302685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1260, + "weighted_flow_matching_loss": 0.02737160213291645, + "weighted_rkd_loss": 0.015501476731151343 + }, + { + "epoch": 2.1709401709401708, + "flow_matching_loss": 0.026486062444746494, + "grad_norm": 0.2249724268913269, + "learning_rate": 2.672039971101979e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011717427894473076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005194938997738064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028629794716835022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1270, + "weighted_flow_matching_loss": 0.026486062444746494, + "weighted_rkd_loss": 0.014314897358417511 + }, + { + "epoch": 2.1880341880341883, + "flow_matching_loss": 0.025420571491122246, + "grad_norm": 0.27121660113334656, + "learning_rate": 2.665822991330178e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012049693893641234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005436556437052787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029535944014787673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1280, + "weighted_flow_matching_loss": 0.025420571491122246, + "weighted_rkd_loss": 0.014767972007393836 + }, + { + "epoch": 2.2051282051282053, + "flow_matching_loss": 0.025054209306836127, + "grad_norm": 0.3212575316429138, + "learning_rate": 2.6595550146839934e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011498598288744688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0056119068060070275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02860910352319479, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1290, + "weighted_flow_matching_loss": 0.025054209306836127, + "weighted_rkd_loss": 0.014304551761597395 + }, + { + "epoch": 2.2222222222222223, + "flow_matching_loss": 0.02525439951568842, + "grad_norm": 0.26853254437446594, + "learning_rate": 2.6532363153450287e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012216195557266473, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005422738334164024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029855129588395358, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1300, + "weighted_flow_matching_loss": 0.02525439951568842, + "weighted_rkd_loss": 0.014927564794197679 + }, + { + "epoch": 2.2393162393162394, + "flow_matching_loss": 0.021937138680368663, + "grad_norm": 0.1946357786655426, + "learning_rate": 2.6468671697136634e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01110919527709484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005047859624028206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02726624980568886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1310, + "weighted_flow_matching_loss": 0.021937138680368663, + "weighted_rkd_loss": 0.01363312490284443 + }, + { + "epoch": 2.2564102564102564, + "flow_matching_loss": 0.02496455330401659, + "grad_norm": 0.22266526520252228, + "learning_rate": 2.640447856396961e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011567748803645372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050113315926864745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02814682889729738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1320, + "weighted_flow_matching_loss": 0.02496455330401659, + "weighted_rkd_loss": 0.01407341444864869 + }, + { + "epoch": 2.2735042735042734, + "flow_matching_loss": 0.02324515637010336, + "grad_norm": 0.22817914187908173, + "learning_rate": 2.6339786561964805e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012382819782942533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005309685179963708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03007532488554716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1330, + "weighted_flow_matching_loss": 0.02324515637010336, + "weighted_rkd_loss": 0.01503766244277358 + }, + { + "epoch": 2.2905982905982905, + "flow_matching_loss": 0.023643253184854985, + "grad_norm": 0.21733051538467407, + "learning_rate": 2.6274598520959966e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010999999195337295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004985643154941499, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026985641196370126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1340, + "weighted_flow_matching_loss": 0.023643253184854985, + "weighted_rkd_loss": 0.013492820598185063 + }, + { + "epoch": 2.3076923076923075, + "flow_matching_loss": 0.022906559705734252, + "grad_norm": 0.28345248103141785, + "learning_rate": 2.6208917292491185e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011069928156211972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005206426791846752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027346283197402954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1350, + "weighted_flow_matching_loss": 0.022906559705734252, + "weighted_rkd_loss": 0.013673141598701477 + }, + { + "epoch": 2.324786324786325, + "flow_matching_loss": 0.02266085259616375, + "grad_norm": 0.1736982762813568, + "learning_rate": 2.6142745749668176e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011071230052039027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004988636053167283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02713109627366066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1360, + "weighted_flow_matching_loss": 0.02266085259616375, + "weighted_rkd_loss": 0.01356554813683033 + }, + { + "epoch": 2.341880341880342, + "flow_matching_loss": 0.02306621465831995, + "grad_norm": 0.26756277680397034, + "learning_rate": 2.607608678704859e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010481357015669346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004531278763897717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0254939928650856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1370, + "weighted_flow_matching_loss": 0.02306621465831995, + "weighted_rkd_loss": 0.0127469964325428 + }, + { + "epoch": 2.358974358974359, + "flow_matching_loss": 0.02586138006299734, + "grad_norm": 0.24713675677776337, + "learning_rate": 2.600894332051139e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011580483987927437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004711030353792012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027871998213231563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1380, + "weighted_flow_matching_loss": 0.02586138006299734, + "weighted_rkd_loss": 0.013935999106615781 + }, + { + "epoch": 2.376068376068376, + "flow_matching_loss": 0.026022879406809807, + "grad_norm": 0.21917226910591125, + "learning_rate": 2.594131828712932e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011406512837857008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004805121943354607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027618147805333136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1390, + "weighted_flow_matching_loss": 0.026022879406809807, + "weighted_rkd_loss": 0.013809073902666568 + }, + { + "epoch": 2.393162393162393, + "flow_matching_loss": 0.023596177250146864, + "grad_norm": 0.2035856992006302, + "learning_rate": 2.5873214645040414e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010142395226284861, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045007105683907865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024785500764846802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1400, + "weighted_flow_matching_loss": 0.023596177250146864, + "weighted_rkd_loss": 0.012392750382423401 + }, + { + "epoch": 2.41025641025641, + "flow_matching_loss": 0.02610907070338726, + "grad_norm": 0.21590881049633026, + "learning_rate": 2.5804635373318604e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008909029187634587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037137356353923678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02153179394081235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1410, + "weighted_flow_matching_loss": 0.02610907070338726, + "weighted_rkd_loss": 0.010765896970406174 + }, + { + "epoch": 2.427350427350427, + "flow_matching_loss": 0.02329109571874142, + "grad_norm": 0.2922363579273224, + "learning_rate": 2.57355834718434e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011788456933572888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004883810528554022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02846072446554899, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1420, + "weighted_flow_matching_loss": 0.02329109571874142, + "weighted_rkd_loss": 0.014230362232774496 + }, + { + "epoch": 2.4444444444444446, + "flow_matching_loss": 0.024115246534347535, + "grad_norm": 0.2587732970714569, + "learning_rate": 2.566606196116867e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013214311562478542, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005839744093827903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03226836714893579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1430, + "weighted_flow_matching_loss": 0.024115246534347535, + "weighted_rkd_loss": 0.016134183574467897 + }, + { + "epoch": 2.4615384615384617, + "flow_matching_loss": 0.02489748504012823, + "grad_norm": 0.21666760742664337, + "learning_rate": 2.559607388239051e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010760731622576713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004546113940887153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026067577116191388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1440, + "weighted_flow_matching_loss": 0.02489748504012823, + "weighted_rkd_loss": 0.013033788558095694 + }, + { + "epoch": 2.4786324786324787, + "flow_matching_loss": 0.020900059305131436, + "grad_norm": 0.2407768964767456, + "learning_rate": 2.552562229701422e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010831856867298484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004969510156661272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02663322389125824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1450, + "weighted_flow_matching_loss": 0.020900059305131436, + "weighted_rkd_loss": 0.01331661194562912 + }, + { + "epoch": 2.4957264957264957, + "flow_matching_loss": 0.022048553638160228, + "grad_norm": 0.16484962403774261, + "learning_rate": 2.545471028682036e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008939298242330551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003671724908053875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02155032129958272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1460, + "weighted_flow_matching_loss": 0.022048553638160228, + "weighted_rkd_loss": 0.01077516064979136 + }, + { + "epoch": 2.5128205128205128, + "flow_matching_loss": 0.024487999454140662, + "grad_norm": 0.1850469410419464, + "learning_rate": 2.538334095373e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010523739224299788, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00469970719423145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025747185945510863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1470, + "weighted_flow_matching_loss": 0.024487999454140662, + "weighted_rkd_loss": 0.012873592972755431 + }, + { + "epoch": 2.52991452991453, + "flow_matching_loss": 0.023977626115083694, + "grad_norm": 0.26976409554481506, + "learning_rate": 2.531151741966896e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01056251721456647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005071718548424542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026196752954274417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1480, + "weighted_flow_matching_loss": 0.023977626115083694, + "weighted_rkd_loss": 0.013098376477137208 + }, + { + "epoch": 2.547008547008547, + "flow_matching_loss": 0.02404904179275036, + "grad_norm": 0.22599978744983673, + "learning_rate": 2.5239242826431302e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010430255578830839, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004550645523704588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02541115628555417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1490, + "weighted_flow_matching_loss": 0.02404904179275036, + "weighted_rkd_loss": 0.012705578142777086 + }, + { + "epoch": 2.564102564102564, + "flow_matching_loss": 0.026297237910330295, + "grad_norm": 0.2992944121360779, + "learning_rate": 2.5166520335541865e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011783476080745458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052417038241401315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02880865577608347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1500, + "weighted_flow_matching_loss": 0.026297237910330295, + "weighted_rkd_loss": 0.014404327888041735 + }, + { + "epoch": 2.5811965811965814, + "flow_matching_loss": 0.023364758491516112, + "grad_norm": 0.31836071610450745, + "learning_rate": 2.5093353128117996e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010980317369103432, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005202659475617111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02716329451650381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1510, + "weighted_flow_matching_loss": 0.023364758491516112, + "weighted_rkd_loss": 0.013581647258251905 + }, + { + "epoch": 2.5982905982905984, + "flow_matching_loss": 0.022494507301598788, + "grad_norm": 0.2116675227880478, + "learning_rate": 2.501974440473037e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010303075471892954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004565139859914779, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025171290896832944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1520, + "weighted_flow_matching_loss": 0.022494507301598788, + "weighted_rkd_loss": 0.012585645448416472 + }, + { + "epoch": 2.6153846153846154, + "flow_matching_loss": 0.026725856214761735, + "grad_norm": 0.24244584143161774, + "learning_rate": 2.4945697385263015e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011164265777915716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005043569020926952, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027372100763022898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1530, + "weighted_flow_matching_loss": 0.026725856214761735, + "weighted_rkd_loss": 0.013686050381511449 + }, + { + "epoch": 2.6324786324786325, + "flow_matching_loss": 0.02198776612058282, + "grad_norm": 0.22380423545837402, + "learning_rate": 2.487121530877243e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010480466578155757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004462654585950077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025423587858676912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1540, + "weighted_flow_matching_loss": 0.02198776612058282, + "weighted_rkd_loss": 0.012711793929338456 + }, + { + "epoch": 2.6495726495726495, + "flow_matching_loss": 0.02080122269690037, + "grad_norm": 0.24043220281600952, + "learning_rate": 2.4796301433345943e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008758699102327228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038610855117440224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02137848371639848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1550, + "weighted_flow_matching_loss": 0.02080122269690037, + "weighted_rkd_loss": 0.01068924185819924 + }, + { + "epoch": 2.6666666666666665, + "flow_matching_loss": 0.023161404207348822, + "grad_norm": 0.24361535906791687, + "learning_rate": 2.4720959035959148e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011636437475681305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005215252633206546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028488127514719963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1560, + "weighted_flow_matching_loss": 0.023161404207348822, + "weighted_rkd_loss": 0.014244063757359982 + }, + { + "epoch": 2.683760683760684, + "flow_matching_loss": 0.02135396469384432, + "grad_norm": 0.24413393437862396, + "learning_rate": 2.464519141233257e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010051483660936356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042491775238886476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024352144729346038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1570, + "weighted_flow_matching_loss": 0.02135396469384432, + "weighted_rkd_loss": 0.012176072364673019 + }, + { + "epoch": 2.700854700854701, + "flow_matching_loss": 0.022891308926045895, + "grad_norm": 0.2563834488391876, + "learning_rate": 2.456900187678754e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009299258328974247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004032260063104332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02263077683746815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1580, + "weighted_flow_matching_loss": 0.022891308926045895, + "weighted_rkd_loss": 0.011315388418734074 + }, + { + "epoch": 2.717948717948718, + "flow_matching_loss": 0.019574376195669173, + "grad_norm": 0.2142282873392105, + "learning_rate": 2.4492393762101147e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01088754371739924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005167186912149191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02694227434694767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1590, + "weighted_flow_matching_loss": 0.019574376195669173, + "weighted_rkd_loss": 0.013471137173473835 + }, + { + "epoch": 2.735042735042735, + "flow_matching_loss": 0.023829315975308418, + "grad_norm": 0.2677934169769287, + "learning_rate": 2.4415370419360508e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010912666656076908, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004734570253640413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026559903100132943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1600, + "weighted_flow_matching_loss": 0.023829315975308418, + "weighted_rkd_loss": 0.013279951550066471 + }, + { + "epoch": 2.752136752136752, + "flow_matching_loss": 0.022046511620283128, + "grad_norm": 0.18143510818481445, + "learning_rate": 2.4337935217816157e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009173155901953578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004238010593689978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022584322560578584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1610, + "weighted_flow_matching_loss": 0.022046511620283128, + "weighted_rkd_loss": 0.011292161280289292 + }, + { + "epoch": 2.769230769230769, + "flow_matching_loss": 0.0256638266146183, + "grad_norm": 0.26739150285720825, + "learning_rate": 2.4260091544734665e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009626912791281939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045542302541434765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02380805565044284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1620, + "weighted_flow_matching_loss": 0.0256638266146183, + "weighted_rkd_loss": 0.01190402782522142 + }, + { + "epoch": 2.786324786324786, + "flow_matching_loss": 0.02314116433262825, + "grad_norm": 0.27141332626342773, + "learning_rate": 2.4181842805250466e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00963348699733615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044304505223408345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02369742449373007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1630, + "weighted_flow_matching_loss": 0.02314116433262825, + "weighted_rkd_loss": 0.011848712246865034 + }, + { + "epoch": 2.8034188034188032, + "flow_matching_loss": 0.02181963212788105, + "grad_norm": 0.20872725546360016, + "learning_rate": 2.4103192422216923e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009032536670565606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004314871900714934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022379945125430824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1640, + "weighted_flow_matching_loss": 0.02181963212788105, + "weighted_rkd_loss": 0.011189972562715412 + }, + { + "epoch": 2.8205128205128203, + "flow_matching_loss": 0.022701138444244862, + "grad_norm": 0.1778896450996399, + "learning_rate": 2.402414383605658e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010498900059610605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045100216753780845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02550782198086381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1650, + "weighted_flow_matching_loss": 0.022701138444244862, + "weighted_rkd_loss": 0.012753910990431905 + }, + { + "epoch": 2.8376068376068377, + "flow_matching_loss": 0.024045399948954584, + "grad_norm": 0.25071728229522705, + "learning_rate": 2.3944700504610674e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00897721559740603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003891855734400451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021846286952495575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1660, + "weighted_flow_matching_loss": 0.024045399948954584, + "weighted_rkd_loss": 0.010923143476247787 + }, + { + "epoch": 2.8547008547008548, + "flow_matching_loss": 0.02234038049355149, + "grad_norm": 0.20470502972602844, + "learning_rate": 2.3864865902987883e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011603549402207137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051022290484979745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028309327736496925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1670, + "weighted_flow_matching_loss": 0.02234038049355149, + "weighted_rkd_loss": 0.014154663868248462 + }, + { + "epoch": 2.871794871794872, + "flow_matching_loss": 0.023738479986786843, + "grad_norm": 0.30578407645225525, + "learning_rate": 2.378464352341231e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008975940523669124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003991176700219512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02194305770099163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1680, + "weighted_flow_matching_loss": 0.023738479986786843, + "weighted_rkd_loss": 0.010971528850495816 + }, + { + "epoch": 2.888888888888889, + "flow_matching_loss": 0.022554483730345964, + "grad_norm": 0.1951245367527008, + "learning_rate": 2.3704036875070728e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009812219673767686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004091479280032217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023715918976813556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1690, + "weighted_flow_matching_loss": 0.022554483730345964, + "weighted_rkd_loss": 0.011857959488406778 + }, + { + "epoch": 2.905982905982906, + "flow_matching_loss": 0.02242042701691389, + "grad_norm": 0.2974187135696411, + "learning_rate": 2.362304948395906e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010304574249312281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004178128391504288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024787276796996594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1700, + "weighted_flow_matching_loss": 0.02242042701691389, + "weighted_rkd_loss": 0.012393638398498297 + }, + { + "epoch": 2.9230769230769234, + "flow_matching_loss": 0.023922652564942838, + "grad_norm": 0.18772614002227783, + "learning_rate": 2.3541684892728157e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009445326868444681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004097510641440749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022988164238631724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1710, + "weighted_flow_matching_loss": 0.023922652564942838, + "weighted_rkd_loss": 0.011494082119315862 + }, + { + "epoch": 2.9401709401709404, + "flow_matching_loss": 0.026682638563215733, + "grad_norm": 0.2330038994550705, + "learning_rate": 2.345994666052883e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009895666409283877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00440065455622971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024191987700760365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1720, + "weighted_flow_matching_loss": 0.026682638563215733, + "weighted_rkd_loss": 0.012095993850380182 + }, + { + "epoch": 2.9572649572649574, + "flow_matching_loss": 0.02124050781130791, + "grad_norm": 0.19767366349697113, + "learning_rate": 2.337783836285614e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008275231532752513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003853382426314056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020403845328837635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1730, + "weighted_flow_matching_loss": 0.02124050781130791, + "weighted_rkd_loss": 0.010201922664418817 + }, + { + "epoch": 2.9743589743589745, + "flow_matching_loss": 0.02081846250221133, + "grad_norm": 0.307229608297348, + "learning_rate": 2.3295363591393026e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009578318148851395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040630410658195615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02321967724710703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1740, + "weighted_flow_matching_loss": 0.02081846250221133, + "weighted_rkd_loss": 0.011609838623553515 + }, + { + "epoch": 2.9914529914529915, + "flow_matching_loss": 0.0220365384593606, + "grad_norm": 0.19888272881507874, + "learning_rate": 2.3212525953853174e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009073050040751696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003744899039156735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021890999004244804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1750, + "weighted_flow_matching_loss": 0.0220365384593606, + "weighted_rkd_loss": 0.010945499502122402 + }, + { + "epoch": 3.0085470085470085, + "flow_matching_loss": 0.020825703255832197, + "grad_norm": 0.1918894499540329, + "learning_rate": 2.3129329073823206e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009859509905800224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045070439809933305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024226063676178456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1760, + "weighted_flow_matching_loss": 0.020825703255832197, + "weighted_rkd_loss": 0.012113031838089228 + }, + { + "epoch": 3.0256410256410255, + "flow_matching_loss": 0.02404018212109804, + "grad_norm": 0.19095730781555176, + "learning_rate": 2.3045776590604175e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008907624846324324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004010485555045307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021825735550373794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1770, + "weighted_flow_matching_loss": 0.02404018212109804, + "weighted_rkd_loss": 0.010912867775186897 + }, + { + "epoch": 3.0427350427350426, + "flow_matching_loss": 0.02037848560139537, + "grad_norm": 0.23975929617881775, + "learning_rate": 2.296187215905237e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009382487926632165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042326563969254495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022997632250189783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1780, + "weighted_flow_matching_loss": 0.02037848560139537, + "weighted_rkd_loss": 0.011498816125094891 + }, + { + "epoch": 3.0598290598290596, + "flow_matching_loss": 0.02206662083044648, + "grad_norm": 0.14885295927524567, + "learning_rate": 2.2877619449419437e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008644994953647256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035399044398218395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020829894207417966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1790, + "weighted_flow_matching_loss": 0.02206662083044648, + "weighted_rkd_loss": 0.010414947103708983 + }, + { + "epoch": 3.076923076923077, + "flow_matching_loss": 0.020480931922793388, + "grad_norm": 0.20446527004241943, + "learning_rate": 2.2793022147191843e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009374905563890933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042403008788824085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02299011219292879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1800, + "weighted_flow_matching_loss": 0.020480931922793388, + "weighted_rkd_loss": 0.011495056096464396 + }, + { + "epoch": 3.094017094017094, + "flow_matching_loss": 0.021702952776104213, + "grad_norm": 0.19124248623847961, + "learning_rate": 2.2708083952929634e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009283293923363089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004115194361656904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022681782208383083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1810, + "weighted_flow_matching_loss": 0.021702952776104213, + "weighted_rkd_loss": 0.011340891104191541 + }, + { + "epoch": 3.111111111111111, + "flow_matching_loss": 0.023065047152340413, + "grad_norm": 0.16134604811668396, + "learning_rate": 2.2622808582104594e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008247178699821233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038086207583546638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020302978157997132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1820, + "weighted_flow_matching_loss": 0.023065047152340413, + "weighted_rkd_loss": 0.010151489078998566 + }, + { + "epoch": 3.128205128205128, + "flow_matching_loss": 0.020875094179064035, + "grad_norm": 0.21919263899326324, + "learning_rate": 2.2537199764937703e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00895846332423389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00419829033780843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022115217056125402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1830, + "weighted_flow_matching_loss": 0.020875094179064035, + "weighted_rkd_loss": 0.011057608528062701 + }, + { + "epoch": 3.1452991452991452, + "flow_matching_loss": 0.02138959923759103, + "grad_norm": 0.23678432404994965, + "learning_rate": 2.245126124623595e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007867638859897852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003307641646824777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019042919389903545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1840, + "weighted_flow_matching_loss": 0.02138959923759103, + "weighted_rkd_loss": 0.009521459694951773 + }, + { + "epoch": 3.1623931623931623, + "flow_matching_loss": 0.02178165577352047, + "grad_norm": 0.27445220947265625, + "learning_rate": 2.2364996785228547e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008633937686681747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004018479073420167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021286354307085274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1850, + "weighted_flow_matching_loss": 0.02178165577352047, + "weighted_rkd_loss": 0.010643177153542637 + }, + { + "epoch": 3.1794871794871793, + "flow_matching_loss": 0.021973855700343846, + "grad_norm": 0.2308157980442047, + "learning_rate": 2.227841015540248e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007728042174130678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032928862725384533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018748970329761507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1860, + "weighted_flow_matching_loss": 0.021973855700343846, + "weighted_rkd_loss": 0.009374485164880753 + }, + { + "epoch": 3.1965811965811968, + "flow_matching_loss": 0.021395375672727823, + "grad_norm": 0.21695679426193237, + "learning_rate": 2.2191505144337433e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077186715323477985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033147968002595007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018752139899879693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1870, + "weighted_flow_matching_loss": 0.021395375672727823, + "weighted_rkd_loss": 0.009376069949939846 + }, + { + "epoch": 3.213675213675214, + "flow_matching_loss": 0.021534157823771237, + "grad_norm": 0.1653919219970703, + "learning_rate": 2.210428555354013e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00908505842089653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035650128265842795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02173512978479266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1880, + "weighted_flow_matching_loss": 0.021534157823771237, + "weighted_rkd_loss": 0.01086756489239633 + }, + { + "epoch": 3.230769230769231, + "flow_matching_loss": 0.02106862384825945, + "grad_norm": 0.2326592653989792, + "learning_rate": 2.201675519827802e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009765352075919508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004243217711336911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023773921933025122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1890, + "weighted_flow_matching_loss": 0.02106862384825945, + "weighted_rkd_loss": 0.011886960966512561 + }, + { + "epoch": 3.247863247863248, + "flow_matching_loss": 0.021206633374094962, + "grad_norm": 0.19293636083602905, + "learning_rate": 2.1928917907412403e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007537059253081679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003404276235960424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018478394765406848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1900, + "weighted_flow_matching_loss": 0.021206633374094962, + "weighted_rkd_loss": 0.009239197382703424 + }, + { + "epoch": 3.264957264957265, + "flow_matching_loss": 0.02262264396995306, + "grad_norm": 0.20528379082679749, + "learning_rate": 2.1840777523230937e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008054467104375362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038436053320765494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019952539540827274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1910, + "weighted_flow_matching_loss": 0.02262264396995306, + "weighted_rkd_loss": 0.009976269770413637 + }, + { + "epoch": 3.282051282051282, + "flow_matching_loss": 0.022744439914822577, + "grad_norm": 0.18302470445632935, + "learning_rate": 2.1752337901279552e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011205578362569213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004998157313093543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027409313805401327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1920, + "weighted_flow_matching_loss": 0.022744439914822577, + "weighted_rkd_loss": 0.013704656902700663 + }, + { + "epoch": 3.299145299145299, + "flow_matching_loss": 0.021091637015342713, + "grad_norm": 0.27177563309669495, + "learning_rate": 2.166360291019383e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008100183354690671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032902485225349666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019490615092217924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1930, + "weighted_flow_matching_loss": 0.021091637015342713, + "weighted_rkd_loss": 0.009745307546108962 + }, + { + "epoch": 3.316239316239316, + "flow_matching_loss": 0.022644120268523693, + "grad_norm": 0.20682282745838165, + "learning_rate": 2.157457643152973e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009497429803013801, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00464288117364049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02363774077966809, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1940, + "weighted_flow_matching_loss": 0.022644120268523693, + "weighted_rkd_loss": 0.011818870389834046 + }, + { + "epoch": 3.3333333333333335, + "flow_matching_loss": 0.02207537516951561, + "grad_norm": 0.24180874228477478, + "learning_rate": 2.148526235959385e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008376868302002549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032720234477892516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020025760307908057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1950, + "weighted_flow_matching_loss": 0.02207537516951561, + "weighted_rkd_loss": 0.010012880153954028 + }, + { + "epoch": 3.3504273504273505, + "flow_matching_loss": 0.02177107408642769, + "grad_norm": 0.25179964303970337, + "learning_rate": 2.1395664601273016e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00934967128559947, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004402558598667383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023101901076734065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1960, + "weighted_flow_matching_loss": 0.02177107408642769, + "weighted_rkd_loss": 0.011550950538367032 + }, + { + "epoch": 3.3675213675213675, + "flow_matching_loss": 0.021467326395213603, + "grad_norm": 0.20472826063632965, + "learning_rate": 2.130578707586345e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00802918765693903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00377117128809914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01982954628765583, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1970, + "weighted_flow_matching_loss": 0.021467326395213603, + "weighted_rkd_loss": 0.009914773143827915 + }, + { + "epoch": 3.3846153846153846, + "flow_matching_loss": 0.022415131144225598, + "grad_norm": 0.18386898934841156, + "learning_rate": 2.1215633714899264e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008150833193212748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038708670064806936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020172533206641675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1980, + "weighted_flow_matching_loss": 0.022415131144225598, + "weighted_rkd_loss": 0.010086266603320838 + }, + { + "epoch": 3.4017094017094016, + "flow_matching_loss": 0.02110731778666377, + "grad_norm": 0.1889202892780304, + "learning_rate": 2.112520846198053e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00840694373473525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003788764402270317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020602651871740818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1990, + "weighted_flow_matching_loss": 0.02110731778666377, + "weighted_rkd_loss": 0.010301325935870409 + }, + { + "epoch": 3.4188034188034186, + "flow_matching_loss": 0.019866282120347023, + "grad_norm": 0.26760584115982056, + "learning_rate": 2.103451527260074e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011116993194445968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044356569647789005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02666964316740632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2000, + "weighted_flow_matching_loss": 0.019866282120347023, + "weighted_rkd_loss": 0.01333482158370316 + }, + { + "epoch": 3.435897435897436, + "flow_matching_loss": 0.018249011039733885, + "grad_norm": 0.23657004535198212, + "learning_rate": 2.094355811397381e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008862052159383894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038880287203937767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021612132992595436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2010, + "weighted_flow_matching_loss": 0.018249011039733885, + "weighted_rkd_loss": 0.010806066496297718 + }, + { + "epoch": 3.452991452991453, + "flow_matching_loss": 0.022592077776789666, + "grad_norm": 0.16656915843486786, + "learning_rate": 2.0852340964860522e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008196860598400236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035904660122469067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019984187465161084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2020, + "weighted_flow_matching_loss": 0.022592077776789666, + "weighted_rkd_loss": 0.009992093732580542 + }, + { + "epoch": 3.47008547008547, + "flow_matching_loss": 0.021013226732611655, + "grad_norm": 0.2137608677148819, + "learning_rate": 2.076086781539447e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008774524834007025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004004479385912418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021553528774529696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2030, + "weighted_flow_matching_loss": 0.021013226732611655, + "weighted_rkd_loss": 0.010776764387264848 + }, + { + "epoch": 3.4871794871794872, + "flow_matching_loss": 0.019182110857218503, + "grad_norm": 0.2665335237979889, + "learning_rate": 2.066914266690755e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008432052610442043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039183218264952305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020782427024096252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2040, + "weighted_flow_matching_loss": 0.019182110857218503, + "weighted_rkd_loss": 0.010391213512048126 + }, + { + "epoch": 3.5042735042735043, + "flow_matching_loss": 0.02038408424705267, + "grad_norm": 0.15456117689609528, + "learning_rate": 2.0577169531754906e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010282731615006924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004464774415828288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025030237529426812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2050, + "weighted_flow_matching_loss": 0.02038408424705267, + "weighted_rkd_loss": 0.012515118764713406 + }, + { + "epoch": 3.5213675213675213, + "flow_matching_loss": 0.021545465849339963, + "grad_norm": 0.1929696649312973, + "learning_rate": 2.048495243313942e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008824501093477011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036449615377932786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02129396367818117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2060, + "weighted_flow_matching_loss": 0.021545465849339963, + "weighted_rkd_loss": 0.010646981839090585 + }, + { + "epoch": 3.5384615384615383, + "flow_matching_loss": 0.018455321434885263, + "grad_norm": 0.23541535437107086, + "learning_rate": 2.039249540493572e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00999117074534297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003966285148635507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023948626406490803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2070, + "weighted_flow_matching_loss": 0.018455321434885263, + "weighted_rkd_loss": 0.011974313203245402 + }, + { + "epoch": 3.5555555555555554, + "flow_matching_loss": 0.018786207027733325, + "grad_norm": 0.22462044656276703, + "learning_rate": 2.0299802491513755e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00863970723003149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035249478416517375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02080436246469617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2080, + "weighted_flow_matching_loss": 0.018786207027733325, + "weighted_rkd_loss": 0.010402181232348085 + }, + { + "epoch": 3.5726495726495724, + "flow_matching_loss": 0.019480399787425995, + "grad_norm": 0.2128298580646515, + "learning_rate": 2.0206877747561828e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007677402580156922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003288325946778059, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01864313129335642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2090, + "weighted_flow_matching_loss": 0.019480399787425995, + "weighted_rkd_loss": 0.00932156564667821 + }, + { + "epoch": 3.58974358974359, + "flow_matching_loss": 0.01911817295476794, + "grad_norm": 0.23664461076259613, + "learning_rate": 2.011372523790927e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00922760833054781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004206936457194388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022662153095006944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2100, + "weighted_flow_matching_loss": 0.01911817295476794, + "weighted_rkd_loss": 0.011331076547503472 + }, + { + "epoch": 3.606837606837607, + "flow_matching_loss": 0.021250076405704022, + "grad_norm": 0.21973693370819092, + "learning_rate": 2.0020349037348643e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007971875090152025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003988684457726777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019932434521615507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2110, + "weighted_flow_matching_loss": 0.021250076405704022, + "weighted_rkd_loss": 0.009966217260807753 + }, + { + "epoch": 3.623931623931624, + "flow_matching_loss": 0.019656579475849867, + "grad_norm": 0.21634699404239655, + "learning_rate": 1.9926753230457443e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00955271115526557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003916142298839987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023021564818918705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2120, + "weighted_flow_matching_loss": 0.019656579475849867, + "weighted_rkd_loss": 0.011510782409459352 + }, + { + "epoch": 3.641025641025641, + "flow_matching_loss": 0.02276723589748144, + "grad_norm": 0.15986588597297668, + "learning_rate": 1.983294191141948e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075195180252194405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003291030717082322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01833006674423814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2130, + "weighted_flow_matching_loss": 0.02276723589748144, + "weighted_rkd_loss": 0.00916503337211907 + }, + { + "epoch": 3.658119658119658, + "flow_matching_loss": 0.020830532629042865, + "grad_norm": 0.21286843717098236, + "learning_rate": 1.973891918384575e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008494880562648178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003529363591223955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020519124809652567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2140, + "weighted_flow_matching_loss": 0.020830532629042865, + "weighted_rkd_loss": 0.010259562404826283 + }, + { + "epoch": 3.6752136752136755, + "flow_matching_loss": 0.023566334135830404, + "grad_norm": 0.24354811012744904, + "learning_rate": 1.9644689160594945e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008519923128187657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036325100576505066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02067235643044114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2150, + "weighted_flow_matching_loss": 0.023566334135830404, + "weighted_rkd_loss": 0.01033617821522057 + }, + { + "epoch": 3.6923076923076925, + "flow_matching_loss": 0.018900660425424577, + "grad_norm": 0.1920747309923172, + "learning_rate": 1.9550255963593553e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007615871122106909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003055795456748456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018287537898868323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2160, + "weighted_flow_matching_loss": 0.018900660425424577, + "weighted_rkd_loss": 0.009143768949434162 + }, + { + "epoch": 3.7094017094017095, + "flow_matching_loss": 0.01953559760004282, + "grad_norm": 0.19230477511882782, + "learning_rate": 1.9455623723655524e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067968220915645364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031586338998749854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016752277966588736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2170, + "weighted_flow_matching_loss": 0.01953559760004282, + "weighted_rkd_loss": 0.008376138983294368 + }, + { + "epoch": 3.7264957264957266, + "flow_matching_loss": 0.020689549297094344, + "grad_norm": 0.23367641866207123, + "learning_rate": 1.9360796580301606e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008445434179157019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003690579044632614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020581447705626487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2180, + "weighted_flow_matching_loss": 0.020689549297094344, + "weighted_rkd_loss": 0.010290723852813244 + }, + { + "epoch": 3.7435897435897436, + "flow_matching_loss": 0.020847134850919248, + "grad_norm": 0.2316247820854187, + "learning_rate": 1.9265778681578245e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008731900528073312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034508655546233056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020914666634052993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2190, + "weighted_flow_matching_loss": 0.020847134850919248, + "weighted_rkd_loss": 0.010457333317026496 + }, + { + "epoch": 3.7606837606837606, + "flow_matching_loss": 0.02032330334186554, + "grad_norm": 0.17385540902614594, + "learning_rate": 1.9170574183876143e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007317671971395612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032924892380833625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017927832994610073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2200, + "weighted_flow_matching_loss": 0.02032330334186554, + "weighted_rkd_loss": 0.008963916497305036 + }, + { + "epoch": 3.7777777777777777, + "flow_matching_loss": 0.016899863351136447, + "grad_norm": 0.26192307472229004, + "learning_rate": 1.9075187251748452e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007569550909101963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003170735482126474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018309837207198143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2210, + "weighted_flow_matching_loss": 0.016899863351136447, + "weighted_rkd_loss": 0.009154918603599072 + }, + { + "epoch": 3.7948717948717947, + "flow_matching_loss": 0.018195388186722995, + "grad_norm": 0.25618091225624084, + "learning_rate": 1.8979622057728596e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007942789187654852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032951420871540904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019180720765143634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2220, + "weighted_flow_matching_loss": 0.018195388186722995, + "weighted_rkd_loss": 0.009590360382571817 + }, + { + "epoch": 3.8119658119658117, + "flow_matching_loss": 0.02252973020076752, + "grad_norm": 0.1578928828239441, + "learning_rate": 1.8883882782147748e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074423910584300755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003065627347677946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017950409650802614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2230, + "weighted_flow_matching_loss": 0.02252973020076752, + "weighted_rkd_loss": 0.008975204825401307 + }, + { + "epoch": 3.8290598290598292, + "flow_matching_loss": 0.02154372911900282, + "grad_norm": 0.20821666717529297, + "learning_rate": 1.8787973612951967e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00786780584603548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032656229101121425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01900123441591859, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2240, + "weighted_flow_matching_loss": 0.02154372911900282, + "weighted_rkd_loss": 0.009500617207959295 + }, + { + "epoch": 3.8461538461538463, + "flow_matching_loss": 0.020391280949115752, + "grad_norm": 0.17726753652095795, + "learning_rate": 1.8691898745519028e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007469217339530587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003028665063902736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01796709969639778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2250, + "weighted_flow_matching_loss": 0.020391280949115752, + "weighted_rkd_loss": 0.00898354984819889 + }, + { + "epoch": 3.8632478632478633, + "flow_matching_loss": 0.02223206451162696, + "grad_norm": 0.21495142579078674, + "learning_rate": 1.8595662382474863e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007301053707487881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003117244434542954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017719351686537266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2260, + "weighted_flow_matching_loss": 0.02223206451162696, + "weighted_rkd_loss": 0.008859675843268633 + }, + { + "epoch": 3.8803418803418803, + "flow_matching_loss": 0.01902555041015148, + "grad_norm": 0.20775483548641205, + "learning_rate": 1.8499268733509766e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073735327925533055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003244478744454682, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017991544399410486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2270, + "weighted_flow_matching_loss": 0.01902555041015148, + "weighted_rkd_loss": 0.008995772199705243 + }, + { + "epoch": 3.8974358974358974, + "flow_matching_loss": 0.021454056072980166, + "grad_norm": 0.19884011149406433, + "learning_rate": 1.84027220151942e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007698776153847575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034163655480369927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018813917879015207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2280, + "weighted_flow_matching_loss": 0.021454056072980166, + "weighted_rkd_loss": 0.009406958939507604 + }, + { + "epoch": 3.9145299145299144, + "flow_matching_loss": 0.020456366799771787, + "grad_norm": 0.21196779608726501, + "learning_rate": 1.83060264507944e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007814684696495534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031296714674681424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018759041093289852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2290, + "weighted_flow_matching_loss": 0.020456366799771787, + "weighted_rkd_loss": 0.009379520546644926 + }, + { + "epoch": 3.931623931623932, + "flow_matching_loss": 0.019748034421354532, + "grad_norm": 0.21080656349658966, + "learning_rate": 1.82091862700876e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007885579485446215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030657083028927446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018836867436766623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2300, + "weighted_flow_matching_loss": 0.019748034421354532, + "weighted_rkd_loss": 0.009418433718383312 + }, + { + "epoch": 3.948717948717949, + "flow_matching_loss": 0.021026203967630865, + "grad_norm": 0.19430294632911682, + "learning_rate": 1.8112205709177023e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007923238072544336, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032659520395100115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019112428463995457, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2310, + "weighted_flow_matching_loss": 0.021026203967630865, + "weighted_rkd_loss": 0.009556214231997729 + }, + { + "epoch": 3.965811965811966, + "flow_matching_loss": 0.019944813381880522, + "grad_norm": 0.32070183753967285, + "learning_rate": 1.8015089010306585e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009109126357361674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003980800183489919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022199052944779397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2320, + "weighted_flow_matching_loss": 0.019944813381880522, + "weighted_rkd_loss": 0.011099526472389699 + }, + { + "epoch": 3.982905982905983, + "flow_matching_loss": 0.019468062836676837, + "grad_norm": 0.25614234805107117, + "learning_rate": 1.7917840421675298e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008891282556578516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003779213014058769, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021561777964234353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2330, + "weighted_flow_matching_loss": 0.019468062836676837, + "weighted_rkd_loss": 0.010780888982117176 + }, + { + "epoch": 4.0, + "flow_matching_loss": 0.017012250050902365, + "grad_norm": 0.16915491223335266, + "learning_rate": 1.7820464197251474e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006800329172983765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029427450615912678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01654340336099267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2340, + "weighted_flow_matching_loss": 0.017012250050902365, + "weighted_rkd_loss": 0.008271701680496336 + }, + { + "epoch": 4.017094017094017, + "flow_matching_loss": 0.023955489788204432, + "grad_norm": 0.22803570330142975, + "learning_rate": 1.7722964596586627e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006895763799548149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002875331451650709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016666858922690153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2350, + "weighted_flow_matching_loss": 0.023955489788204432, + "weighted_rkd_loss": 0.008333429461345077 + }, + { + "epoch": 4.034188034188034, + "flow_matching_loss": 0.017703853361308575, + "grad_norm": 0.19360429048538208, + "learning_rate": 1.7625345884629144e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007683675503358245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003063351823948324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018430702853947878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2360, + "weighted_flow_matching_loss": 0.017703853361308575, + "weighted_rkd_loss": 0.009215351426973939 + }, + { + "epoch": 4.051282051282051, + "flow_matching_loss": 0.02146550826728344, + "grad_norm": 0.15965017676353455, + "learning_rate": 1.7527612331537722e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066555246710777284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002755923569202423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016066972725093366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2370, + "weighted_flow_matching_loss": 0.02146550826728344, + "weighted_rkd_loss": 0.008033486362546683 + }, + { + "epoch": 4.068376068376068, + "flow_matching_loss": 0.01993013881146908, + "grad_norm": 0.20406599342823029, + "learning_rate": 1.7429768212494593e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006303849536925555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027177043608389794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015325403492897748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2380, + "weighted_flow_matching_loss": 0.01993013881146908, + "weighted_rkd_loss": 0.007662701746448874 + }, + { + "epoch": 4.085470085470085, + "flow_matching_loss": 0.01773031707853079, + "grad_norm": 0.19607993960380554, + "learning_rate": 1.7331817807518497e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007189767109230161, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032816751045174897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017661209404468536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2390, + "weighted_flow_matching_loss": 0.01773031707853079, + "weighted_rkd_loss": 0.008830604702234268 + }, + { + "epoch": 4.102564102564102, + "flow_matching_loss": 0.018448451347649098, + "grad_norm": 0.1936192661523819, + "learning_rate": 1.723376540127745e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006861878465861082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031465153908357022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016870272066444158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2400, + "weighted_flow_matching_loss": 0.018448451347649098, + "weighted_rkd_loss": 0.008435136033222079 + }, + { + "epoch": 4.119658119658119, + "flow_matching_loss": 0.01903025442734361, + "grad_norm": 0.28160524368286133, + "learning_rate": 1.7135615282901357e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007035943772643804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002800643094815314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016872530709952115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2410, + "weighted_flow_matching_loss": 0.01903025442734361, + "weighted_rkd_loss": 0.008436265354976057 + }, + { + "epoch": 4.136752136752137, + "flow_matching_loss": 0.02104166429489851, + "grad_norm": 0.15420423448085785, + "learning_rate": 1.7037371745794358e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008220721362158656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032787902280688288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019720232859253884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2420, + "weighted_flow_matching_loss": 0.02104166429489851, + "weighted_rkd_loss": 0.009860116429626942 + }, + { + "epoch": 4.153846153846154, + "flow_matching_loss": 0.019651689007878302, + "grad_norm": 0.15077415108680725, + "learning_rate": 1.6939039087447026e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075453647645190355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031900717760436235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018280801270157097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2430, + "weighted_flow_matching_loss": 0.019651689007878302, + "weighted_rkd_loss": 0.009140400635078549 + }, + { + "epoch": 4.170940170940171, + "flow_matching_loss": 0.020570564549416304, + "grad_norm": 0.15573781728744507, + "learning_rate": 1.6840621609248397e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007544708624482155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003139568492770195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018228985741734503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2440, + "weighted_flow_matching_loss": 0.020570564549416304, + "weighted_rkd_loss": 0.009114492870867252 + }, + { + "epoch": 4.188034188034188, + "flow_matching_loss": 0.019456946291029453, + "grad_norm": 0.20012281835079193, + "learning_rate": 1.6742123616297794e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006495161354541779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027588428696617485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01574916560202837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2450, + "weighted_flow_matching_loss": 0.019456946291029453, + "weighted_rkd_loss": 0.007874582801014185 + }, + { + "epoch": 4.205128205128205, + "flow_matching_loss": 0.019614775478839875, + "grad_norm": 0.1647314727306366, + "learning_rate": 1.664354941721651e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006208974169567227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002630527503788471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01504847602918744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2460, + "weighted_flow_matching_loss": 0.019614775478839875, + "weighted_rkd_loss": 0.00752423801459372 + }, + { + "epoch": 4.222222222222222, + "flow_matching_loss": 0.0180529972538352, + "grad_norm": 0.1961066573858261, + "learning_rate": 1.6544903323959355e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073069647885859014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002847653639037162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017461583390831946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2470, + "weighted_flow_matching_loss": 0.0180529972538352, + "weighted_rkd_loss": 0.008730791695415973 + }, + { + "epoch": 4.239316239316239, + "flow_matching_loss": 0.020445212721824646, + "grad_norm": 0.22484345734119415, + "learning_rate": 1.6446189651626023e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006985000707209111, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031812989618629217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017151300329715014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2480, + "weighted_flow_matching_loss": 0.020445212721824646, + "weighted_rkd_loss": 0.008575650164857507 + }, + { + "epoch": 4.256410256410256, + "flow_matching_loss": 0.019169483799487352, + "grad_norm": 0.2385309785604477, + "learning_rate": 1.6347412718272316e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006840714206919074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025678937672637403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016249322239309548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2490, + "weighted_flow_matching_loss": 0.019169483799487352, + "weighted_rkd_loss": 0.008124661119654774 + }, + { + "epoch": 4.273504273504273, + "flow_matching_loss": 0.017080207541584968, + "grad_norm": 0.17671377956867218, + "learning_rate": 1.624857684472131e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007922574994154274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002863072499167174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018708222638815643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2500, + "weighted_flow_matching_loss": 0.017080207541584968, + "weighted_rkd_loss": 0.009354111319407821 + }, + { + "epoch": 4.2905982905982905, + "flow_matching_loss": 0.019747746549546717, + "grad_norm": 0.24573016166687012, + "learning_rate": 1.6149686354374283e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007329531060531735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030151148559525607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017674177046865226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2510, + "weighted_flow_matching_loss": 0.019747746549546717, + "weighted_rkd_loss": 0.008837088523432613 + }, + { + "epoch": 4.3076923076923075, + "flow_matching_loss": 0.018726011179387568, + "grad_norm": 0.22373729944229126, + "learning_rate": 1.605074557302165e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006631332472898066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002862296684179455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016124961618334054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2520, + "weighted_flow_matching_loss": 0.018726011179387568, + "weighted_rkd_loss": 0.008062480809167027 + }, + { + "epoch": 4.3247863247863245, + "flow_matching_loss": 0.021509300265461207, + "grad_norm": 0.1717628836631775, + "learning_rate": 1.59517588286537e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006188838463276625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025393213145434856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01491699842736125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2530, + "weighted_flow_matching_loss": 0.021509300265461207, + "weighted_rkd_loss": 0.007458499213680625 + }, + { + "epoch": 4.3418803418803416, + "flow_matching_loss": 0.020329602248966693, + "grad_norm": 0.18526530265808105, + "learning_rate": 1.5852730451271324e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007229596329852939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002875377354212105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01733456999063492, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2540, + "weighted_flow_matching_loss": 0.020329602248966693, + "weighted_rkd_loss": 0.00866728499531746 + }, + { + "epoch": 4.358974358974359, + "flow_matching_loss": 0.0184131451882422, + "grad_norm": 0.18414761126041412, + "learning_rate": 1.5753664772696546e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006622652034275233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026653855573385956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01591068971902132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2550, + "weighted_flow_matching_loss": 0.0184131451882422, + "weighted_rkd_loss": 0.00795534485951066 + }, + { + "epoch": 4.3760683760683765, + "flow_matching_loss": 0.02079639583826065, + "grad_norm": 0.18560418486595154, + "learning_rate": 1.565456612638307e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007964245975017548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003640728979371488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019569220952689648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2560, + "weighted_flow_matching_loss": 0.02079639583826065, + "weighted_rkd_loss": 0.009784610476344824 + }, + { + "epoch": 4.3931623931623935, + "flow_matching_loss": 0.01935685221105814, + "grad_norm": 0.20377367734909058, + "learning_rate": 1.5555438847226732e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062549715396016834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029534086119383574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015463351644575596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2570, + "weighted_flow_matching_loss": 0.01935685221105814, + "weighted_rkd_loss": 0.007731675822287798 + }, + { + "epoch": 4.410256410256411, + "flow_matching_loss": 0.018714569322764874, + "grad_norm": 0.16832947731018066, + "learning_rate": 1.545628727137584e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006530326465144754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027387462090700866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015799399092793466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2580, + "weighted_flow_matching_loss": 0.018714569322764874, + "weighted_rkd_loss": 0.007899699546396733 + }, + { + "epoch": 4.427350427350428, + "flow_matching_loss": 0.019943490531295537, + "grad_norm": 0.17356619238853455, + "learning_rate": 1.535711573604153e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006290619866922498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028495809994637964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015430820826441049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2590, + "weighted_flow_matching_loss": 0.019943490531295537, + "weighted_rkd_loss": 0.0077154104132205244 + }, + { + "epoch": 4.444444444444445, + "flow_matching_loss": 0.017166621424257754, + "grad_norm": 0.1952560544013977, + "learning_rate": 1.5257928579308032e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006391754373908043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028482529683969914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015631761588156223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2600, + "weighted_flow_matching_loss": 0.017166621424257754, + "weighted_rkd_loss": 0.007815880794078111 + }, + { + "epoch": 4.461538461538462, + "flow_matching_loss": 0.017560683749616147, + "grad_norm": 0.1988353580236435, + "learning_rate": 1.5158730139942902e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007574155880138278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033369460492394865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018485257867723703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2610, + "weighted_flow_matching_loss": 0.017560683749616147, + "weighted_rkd_loss": 0.009242628933861852 + }, + { + "epoch": 4.478632478632479, + "flow_matching_loss": 0.019688459113240242, + "grad_norm": 0.19683468341827393, + "learning_rate": 1.5059524757207239e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063693718519061806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031602281494997444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015898971911519767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2620, + "weighted_flow_matching_loss": 0.019688459113240242, + "weighted_rkd_loss": 0.007949485955759884 + }, + { + "epoch": 4.495726495726496, + "flow_matching_loss": 0.01867778431624174, + "grad_norm": 0.2284543812274933, + "learning_rate": 1.4960316770665866e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006279498664662242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026411876315250993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015200185030698777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2630, + "weighted_flow_matching_loss": 0.01867778431624174, + "weighted_rkd_loss": 0.007600092515349388 + }, + { + "epoch": 4.512820512820513, + "flow_matching_loss": 0.01726768594235182, + "grad_norm": 0.1803433746099472, + "learning_rate": 1.4861110519997495e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005782140232622623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024866451509296893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014050925616174936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2640, + "weighted_flow_matching_loss": 0.01726768594235182, + "weighted_rkd_loss": 0.007025462808087468 + }, + { + "epoch": 4.52991452991453, + "flow_matching_loss": 0.018991167843341827, + "grad_norm": 0.24880416691303253, + "learning_rate": 1.4761910344804913e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008187596360221506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00336456069489941, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01973975347355008, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2650, + "weighted_flow_matching_loss": 0.018991167843341827, + "weighted_rkd_loss": 0.00986987673677504 + }, + { + "epoch": 4.547008547008547, + "flow_matching_loss": 0.020480296481400727, + "grad_norm": 0.15083114802837372, + "learning_rate": 1.466272058442516e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006214395631104708, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026107010897248985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015039492212235928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2660, + "weighted_flow_matching_loss": 0.020480296481400727, + "weighted_rkd_loss": 0.007519746106117964 + }, + { + "epoch": 4.564102564102564, + "flow_matching_loss": 0.016016822867095472, + "grad_norm": 0.24289827048778534, + "learning_rate": 1.4563545577739676e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007098955847322941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003096250956878066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017294162698090077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2670, + "weighted_flow_matching_loss": 0.016016822867095472, + "weighted_rkd_loss": 0.008647081349045039 + }, + { + "epoch": 4.581196581196581, + "flow_matching_loss": 0.016501933615654708, + "grad_norm": 0.23007218539714813, + "learning_rate": 1.4464389662984543e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073565777856856585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026881229132413864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017401278391480444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2680, + "weighted_flow_matching_loss": 0.016501933615654708, + "weighted_rkd_loss": 0.008700639195740222 + }, + { + "epoch": 4.598290598290598, + "flow_matching_loss": 0.016896218247711658, + "grad_norm": 0.2506307065486908, + "learning_rate": 1.4365257177560684e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00605305025819689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002633159328252077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014739259891211987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2690, + "weighted_flow_matching_loss": 0.016896218247711658, + "weighted_rkd_loss": 0.007369629945605994 + }, + { + "epoch": 4.615384615384615, + "flow_matching_loss": 0.019083459116518497, + "grad_norm": 0.1666194647550583, + "learning_rate": 1.426615245784416e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006203994457609952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027952198404818775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015203209035098553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2700, + "weighted_flow_matching_loss": 0.019083459116518497, + "weighted_rkd_loss": 0.0076016045175492765 + }, + { + "epoch": 4.632478632478632, + "flow_matching_loss": 0.017737826984375715, + "grad_norm": 0.20131826400756836, + "learning_rate": 1.4167079838996468e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005915038753300905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021665282547473907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013996605761349201, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2710, + "weighted_flow_matching_loss": 0.017737826984375715, + "weighted_rkd_loss": 0.006998302880674601 + }, + { + "epoch": 4.64957264957265, + "flow_matching_loss": 0.01762369042262435, + "grad_norm": 0.2097741961479187, + "learning_rate": 1.4068043654774897e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066073834663257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024939873139373956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0157087542116642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2720, + "weighted_flow_matching_loss": 0.01762369042262435, + "weighted_rkd_loss": 0.0078543771058321 + }, + { + "epoch": 4.666666666666667, + "flow_matching_loss": 0.017110300622880457, + "grad_norm": 0.1796332746744156, + "learning_rate": 1.3969048237342992e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006355493422597646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002593577420338988, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015304564125835895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2730, + "weighted_flow_matching_loss": 0.017110300622880457, + "weighted_rkd_loss": 0.007652282062917948 + }, + { + "epoch": 4.683760683760684, + "flow_matching_loss": 0.020454686228185893, + "grad_norm": 0.18637192249298096, + "learning_rate": 1.3870097917081011e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053040543105453255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002318077115342021, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012926185876131058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2740, + "weighted_flow_matching_loss": 0.020454686228185893, + "weighted_rkd_loss": 0.006463092938065529 + }, + { + "epoch": 4.700854700854701, + "flow_matching_loss": 0.016615460719913246, + "grad_norm": 0.18428291380405426, + "learning_rate": 1.3771197022396508e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006608629692345857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031284614000469447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016345720645040272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2750, + "weighted_flow_matching_loss": 0.016615460719913246, + "weighted_rkd_loss": 0.008172860322520136 + }, + { + "epoch": 4.717948717948718, + "flow_matching_loss": 0.018700057454407216, + "grad_norm": 0.15285761654376984, + "learning_rate": 1.3672349879535015e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008110217843204736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003631356067489833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019851791579276322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2760, + "weighted_flow_matching_loss": 0.018700057454407216, + "weighted_rkd_loss": 0.009925895789638161 + }, + { + "epoch": 4.735042735042735, + "flow_matching_loss": 0.016846880689263342, + "grad_norm": 0.18228141963481903, + "learning_rate": 1.3573560812390781e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005839766724966467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002764455007854849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01444398844614625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2770, + "weighted_flow_matching_loss": 0.016846880689263342, + "weighted_rkd_loss": 0.007221994223073125 + }, + { + "epoch": 4.752136752136752, + "flow_matching_loss": 0.017821814958006145, + "grad_norm": 0.17568399012088776, + "learning_rate": 1.3474834142317635e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006056840531527996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024255528813228013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014539233967661857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2780, + "weighted_flow_matching_loss": 0.017821814958006145, + "weighted_rkd_loss": 0.007269616983830929 + }, + { + "epoch": 4.769230769230769, + "flow_matching_loss": 0.01895545031875372, + "grad_norm": 0.12359953671693802, + "learning_rate": 1.3376174187939955e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006296288664452731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026273989700712264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015219976473599673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2790, + "weighted_flow_matching_loss": 0.01895545031875372, + "weighted_rkd_loss": 0.0076099882367998365 + }, + { + "epoch": 4.786324786324786, + "flow_matching_loss": 0.01891577858477831, + "grad_norm": 0.15576958656311035, + "learning_rate": 1.3277585264963761e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005916955787688494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002357321046292782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014191232435405255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2800, + "weighted_flow_matching_loss": 0.01891577858477831, + "weighted_rkd_loss": 0.0070956162177026275 + }, + { + "epoch": 4.803418803418803, + "flow_matching_loss": 0.018333879485726356, + "grad_norm": 0.22275856137275696, + "learning_rate": 1.3179071685987928e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006870065536350012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026655447203665973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016405675560235977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2810, + "weighted_flow_matching_loss": 0.018333879485726356, + "weighted_rkd_loss": 0.008202837780117989 + }, + { + "epoch": 4.82051282051282, + "flow_matching_loss": 0.019350035302340984, + "grad_norm": 0.18743370473384857, + "learning_rate": 1.308063776031554e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00618456348311156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021372218267060815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014506348874419928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2820, + "weighted_flow_matching_loss": 0.019350035302340984, + "weighted_rkd_loss": 0.007253174437209964 + }, + { + "epoch": 4.837606837606837, + "flow_matching_loss": 0.016245046723634006, + "grad_norm": 0.20033948123455048, + "learning_rate": 1.2982287793765389e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066329317167401316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023788953665643932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015644758567214014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2830, + "weighted_flow_matching_loss": 0.016245046723634006, + "weighted_rkd_loss": 0.007822379283607007 + }, + { + "epoch": 4.854700854700854, + "flow_matching_loss": 0.016420507896691562, + "grad_norm": 0.1553906500339508, + "learning_rate": 1.2884026088483616e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058491195552051066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002730893774423748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014429132873192429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2840, + "weighted_flow_matching_loss": 0.016420507896691562, + "weighted_rkd_loss": 0.007214566436596214 + }, + { + "epoch": 4.871794871794872, + "flow_matching_loss": 0.021777248568832876, + "grad_norm": 0.17237120866775513, + "learning_rate": 1.2785856942755543e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005849478789605201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024362468044273554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014135204302147032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2850, + "weighted_flow_matching_loss": 0.021777248568832876, + "weighted_rkd_loss": 0.007067602151073516 + }, + { + "epoch": 4.888888888888889, + "flow_matching_loss": 0.018680552765727044, + "grad_norm": 0.19036276638507843, + "learning_rate": 1.2687784650817616e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00545023640152067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002043917903210968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012944390717893838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2860, + "weighted_flow_matching_loss": 0.018680552765727044, + "weighted_rkd_loss": 0.006472195358946919 + }, + { + "epoch": 4.905982905982906, + "flow_matching_loss": 0.0194564713165164, + "grad_norm": 0.18401499092578888, + "learning_rate": 1.2589813502669603e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007179317437112332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028180433670058846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01717667831107974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2870, + "weighted_flow_matching_loss": 0.0194564713165164, + "weighted_rkd_loss": 0.00858833915553987 + }, + { + "epoch": 4.923076923076923, + "flow_matching_loss": 0.018249529600143432, + "grad_norm": 0.15625852346420288, + "learning_rate": 1.2491947783886907e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005985748581588268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025901356828399004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014561632834374904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2880, + "weighted_flow_matching_loss": 0.018249529600143432, + "weighted_rkd_loss": 0.007280816417187452 + }, + { + "epoch": 4.94017094017094, + "flow_matching_loss": 0.019586964324116708, + "grad_norm": 0.19170287251472473, + "learning_rate": 1.2394191775433106e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006269854912534356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026000387500971557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015139748714864254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2890, + "weighted_flow_matching_loss": 0.019586964324116708, + "weighted_rkd_loss": 0.007569874357432127 + }, + { + "epoch": 4.957264957264957, + "flow_matching_loss": 0.018550141993910075, + "grad_norm": 0.217694953083992, + "learning_rate": 1.2296549753472696e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063949782401323315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023791401297785343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015169096365571021, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2900, + "weighted_flow_matching_loss": 0.018550141993910075, + "weighted_rkd_loss": 0.007584548182785511 + }, + { + "epoch": 4.9743589743589745, + "flow_matching_loss": 0.01646477123722434, + "grad_norm": 0.1519484668970108, + "learning_rate": 1.2199025989184032e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006408982444554567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00265109499450773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015469059906899929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2910, + "weighted_flow_matching_loss": 0.01646477123722434, + "weighted_rkd_loss": 0.007734529953449964 + }, + { + "epoch": 4.9914529914529915, + "flow_matching_loss": 0.01712658563628793, + "grad_norm": 0.17936837673187256, + "learning_rate": 1.21016247485725e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005534190079197287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002666365006007254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01373474532738328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2920, + "weighted_flow_matching_loss": 0.01712658563628793, + "weighted_rkd_loss": 0.00686737266369164 + }, + { + "epoch": 5.0085470085470085, + "flow_matching_loss": 0.017433699034154416, + "grad_norm": 0.19425033032894135, + "learning_rate": 1.2004350292283897e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005776865407824516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002266028500162065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013819759339094162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2930, + "weighted_flow_matching_loss": 0.017433699034154416, + "weighted_rkd_loss": 0.006909879669547081 + }, + { + "epoch": 5.0256410256410255, + "flow_matching_loss": 0.01623713644221425, + "grad_norm": 0.15402722358703613, + "learning_rate": 1.1907206875418068e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005933634308166802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025461736135184766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01441344222985208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2940, + "weighted_flow_matching_loss": 0.01623713644221425, + "weighted_rkd_loss": 0.00720672111492604 + }, + { + "epoch": 5.042735042735043, + "flow_matching_loss": 0.01722178068011999, + "grad_norm": 0.17161601781845093, + "learning_rate": 1.1810198747342769e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006133737275376916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025711448513902726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014838619343936443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2950, + "weighted_flow_matching_loss": 0.01722178068011999, + "weighted_rkd_loss": 0.007419309671968221 + }, + { + "epoch": 5.05982905982906, + "flow_matching_loss": 0.01563998917117715, + "grad_norm": 0.17408081889152527, + "learning_rate": 1.1713330151507786e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006273146439343691, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002796279580798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01534257223829627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2960, + "weighted_flow_matching_loss": 0.01563998917117715, + "weighted_rkd_loss": 0.007671286119148135 + }, + { + "epoch": 5.076923076923077, + "flow_matching_loss": 0.018053919449448584, + "grad_norm": 0.18406715989112854, + "learning_rate": 1.161660532525931e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006768700829707086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002817504294216633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01635490609332919, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2970, + "weighted_flow_matching_loss": 0.018053919449448584, + "weighted_rkd_loss": 0.008177453046664595 + }, + { + "epoch": 5.094017094017094, + "flow_matching_loss": 0.015317833703011274, + "grad_norm": 0.1968030482530594, + "learning_rate": 1.1520028499654584e-05, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00502970430534333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002209481818135828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01226889044046402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2980, + "weighted_flow_matching_loss": 0.015317833703011274, + "weighted_rkd_loss": 0.00613444522023201 + }, + { + "epoch": 5.111111111111111, + "flow_matching_loss": 0.016654190700501205, + "grad_norm": 0.1656157225370407, + "learning_rate": 1.1423603899276833e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005547607317566872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024255677242763342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013520782347768545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2990, + "weighted_flow_matching_loss": 0.016654190700501205, + "weighted_rkd_loss": 0.006760391173884273 + }, + { + "epoch": 5.128205128205128, + "flow_matching_loss": 0.018000909499824046, + "grad_norm": 0.1773122400045395, + "learning_rate": 1.1327335742050442e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006210618140175939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002675327961333096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01509656421840191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3000, + "weighted_flow_matching_loss": 0.018000909499824046, + "weighted_rkd_loss": 0.007548282109200955 + }, + { + "epoch": 5.145299145299146, + "flow_matching_loss": 0.017652605939656497, + "grad_norm": 0.19575658440589905, + "learning_rate": 1.1231228239056467e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0056807628832757475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025271081482060253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013888633996248245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3010, + "weighted_flow_matching_loss": 0.017652605939656497, + "weighted_rkd_loss": 0.006944316998124122 + }, + { + "epoch": 5.162393162393163, + "flow_matching_loss": 0.01770560061559081, + "grad_norm": 0.15567773580551147, + "learning_rate": 1.113528559434845e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005423327488824725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022221908904612063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013068845868110657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3020, + "weighted_flow_matching_loss": 0.01770560061559081, + "weighted_rkd_loss": 0.0065344229340553285 + }, + { + "epoch": 5.17948717948718, + "flow_matching_loss": 0.01779159056022763, + "grad_norm": 0.14029021561145782, + "learning_rate": 1.1039512004768476e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005052718450315297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002226500539109111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012331937439739704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3030, + "weighted_flow_matching_loss": 0.01779159056022763, + "weighted_rkd_loss": 0.006165968719869852 + }, + { + "epoch": 5.196581196581197, + "flow_matching_loss": 0.01728218114003539, + "grad_norm": 0.15812712907791138, + "learning_rate": 1.0943911659763612e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005331483390182257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002207448345143348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012870415206998586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3040, + "weighted_flow_matching_loss": 0.01728218114003539, + "weighted_rkd_loss": 0.006435207603499293 + }, + { + "epoch": 5.213675213675214, + "flow_matching_loss": 0.017938074190169573, + "grad_norm": 0.1635051965713501, + "learning_rate": 1.084848874120266e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005749216396361589, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002487788395956159, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013986221328377724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3050, + "weighted_flow_matching_loss": 0.017938074190169573, + "weighted_rkd_loss": 0.006993110664188862 + }, + { + "epoch": 5.230769230769231, + "flow_matching_loss": 0.01776145091280341, + "grad_norm": 0.21638160943984985, + "learning_rate": 1.0753247423193196e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0055127232801169155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022512326017022134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01327667897567153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3060, + "weighted_flow_matching_loss": 0.01776145091280341, + "weighted_rkd_loss": 0.006638339487835765 + }, + { + "epoch": 5.247863247863248, + "flow_matching_loss": 0.016698889154940843, + "grad_norm": 0.1955612450838089, + "learning_rate": 1.0658191871899012e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005643902649171651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020833177142776547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01337112309411168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3070, + "weighted_flow_matching_loss": 0.016698889154940843, + "weighted_rkd_loss": 0.00668556154705584 + }, + { + "epoch": 5.264957264957265, + "flow_matching_loss": 0.018136249482631685, + "grad_norm": 0.18112897872924805, + "learning_rate": 1.0563326245357852e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006050816411152482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026447852607816456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014746417943388223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3080, + "weighted_flow_matching_loss": 0.018136249482631685, + "weighted_rkd_loss": 0.0073732089716941115 + }, + { + "epoch": 5.282051282051282, + "flow_matching_loss": 0.016763995122164487, + "grad_norm": 0.14852160215377808, + "learning_rate": 1.0468654693299557e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006076430296525359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00255180976819247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014704670384526252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3090, + "weighted_flow_matching_loss": 0.016763995122164487, + "weighted_rkd_loss": 0.007352335192263126 + }, + { + "epoch": 5.299145299145299, + "flow_matching_loss": 0.01793050765991211, + "grad_norm": 0.14685483276844025, + "learning_rate": 1.0374181356964501e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005478386953473091, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002295808936469257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013252582773566246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3100, + "weighted_flow_matching_loss": 0.01793050765991211, + "weighted_rkd_loss": 0.006626291386783123 + }, + { + "epoch": 5.316239316239316, + "flow_matching_loss": 0.016466488875448704, + "grad_norm": 0.18847137689590454, + "learning_rate": 1.0279910368922465e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005654708738438785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002288639184553176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013598056323826314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3110, + "weighted_flow_matching_loss": 0.016466488875448704, + "weighted_rkd_loss": 0.006799028161913157 + }, + { + "epoch": 5.333333333333333, + "flow_matching_loss": 0.01684662764891982, + "grad_norm": 0.14454138278961182, + "learning_rate": 1.018584585289186e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005896137282252312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002443724067416042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014235998783260585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3120, + "weighted_flow_matching_loss": 0.01684662764891982, + "weighted_rkd_loss": 0.007117999391630292 + }, + { + "epoch": 5.35042735042735, + "flow_matching_loss": 0.019074271619319915, + "grad_norm": 0.2391997128725052, + "learning_rate": 1.0091991923559348e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058216626523062585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024093637708574533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01405268907546997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3130, + "weighted_flow_matching_loss": 0.019074271619319915, + "weighted_rkd_loss": 0.007026344537734985 + }, + { + "epoch": 5.367521367521368, + "flow_matching_loss": 0.01673125522211194, + "grad_norm": 0.15762180089950562, + "learning_rate": 9.99835268639984e-06, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00576824105810374, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002436055976431817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01397253768518567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3140, + "weighted_flow_matching_loss": 0.01673125522211194, + "weighted_rkd_loss": 0.006986268842592835 + }, + { + "epoch": 5.384615384615385, + "flow_matching_loss": 0.015424938220530749, + "grad_norm": 0.18702711164951324, + "learning_rate": 9.904932237496913e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005303484806790948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022581320954486728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012865101406350731, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3150, + "weighted_flow_matching_loss": 0.015424938220530749, + "weighted_rkd_loss": 0.006432550703175366 + }, + { + "epoch": 5.401709401709402, + "flow_matching_loss": 0.015545273665338755, + "grad_norm": 0.20115116238594055, + "learning_rate": 9.811734663363659e-06, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005540975974872709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002636688796337694, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013718640618026256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3160, + "weighted_flow_matching_loss": 0.015545273665338755, + "weighted_rkd_loss": 0.006859320309013128 + }, + { + "epoch": 5.418803418803419, + "flow_matching_loss": 0.017876550648361446, + "grad_norm": 0.1556462049484253, + "learning_rate": 9.718764040763875e-06, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00610649767331779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003013874718453735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01522686993703246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3170, + "weighted_flow_matching_loss": 0.017876550648361446, + "weighted_rkd_loss": 0.00761343496851623 + }, + { + "epoch": 5.435897435897436, + "flow_matching_loss": 0.015284168068319559, + "grad_norm": 0.2237396538257599, + "learning_rate": 9.626024436533777e-06, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005981934443116188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00246106336126104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01442493237555027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3180, + "weighted_flow_matching_loss": 0.015284168068319559, + "weighted_rkd_loss": 0.007212466187775135 + }, + { + "epoch": 5.452991452991453, + "flow_matching_loss": 0.01647183932363987, + "grad_norm": 0.22805136442184448, + "learning_rate": 9.533519907404095e-06, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005533198988996446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002371192118152976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013437590096145868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3190, + "weighted_flow_matching_loss": 0.01647183932363987, + "weighted_rkd_loss": 0.006718795048072934 + }, + { + "epoch": 5.47008547008547, + "flow_matching_loss": 0.01914637051522732, + "grad_norm": 0.15872231125831604, + "learning_rate": 9.4412544998226e-06, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006629107845947146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002984438475687057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016242654342204333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3200, + "weighted_flow_matching_loss": 0.01914637051522732, + "weighted_rkd_loss": 0.008121327171102167 + }, + { + "epoch": 5.487179487179487, + "flow_matching_loss": 0.017620038706809282, + "grad_norm": 0.1488192230463028, + "learning_rate": 9.349232249777124e-06, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005304252705536782, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021423382102511825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012750843539834023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3210, + "weighted_flow_matching_loss": 0.017620038706809282, + "weighted_rkd_loss": 0.006375421769917011 + }, + { + "epoch": 5.504273504273504, + "flow_matching_loss": 0.020184773299843074, + "grad_norm": 0.1621168553829193, + "learning_rate": 9.257457182618985e-06, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0056879224721342325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002303929906338453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013679774757474662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3220, + "weighted_flow_matching_loss": 0.020184773299843074, + "weighted_rkd_loss": 0.006839887378737331 + }, + { + "epoch": 5.521367521367521, + "flow_matching_loss": 0.016489654313772917, + "grad_norm": 0.17492899298667908, + "learning_rate": 9.165933312886944e-06, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005365319689735771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00215536889154464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012886008247733116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3230, + "weighted_flow_matching_loss": 0.016489654313772917, + "weighted_rkd_loss": 0.006443004123866558 + }, + { + "epoch": 5.538461538461538, + "flow_matching_loss": 0.016707110032439233, + "grad_norm": 0.17003421485424042, + "learning_rate": 9.07466464413155e-06, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005001226905733347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022874437272548676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012289897538721561, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3240, + "weighted_flow_matching_loss": 0.016707110032439233, + "weighted_rkd_loss": 0.006144948769360781 + }, + { + "epoch": 5.555555555555555, + "flow_matching_loss": 0.01858956813812256, + "grad_norm": 0.15823964774608612, + "learning_rate": 8.983655168740048e-06, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005134093156084418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019241061294451357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012192292232066393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3250, + "weighted_flow_matching_loss": 0.01858956813812256, + "weighted_rkd_loss": 0.006096146116033196 + }, + { + "epoch": 5.572649572649572, + "flow_matching_loss": 0.01682747919112444, + "grad_norm": 0.19594405591487885, + "learning_rate": 8.892908867761718e-06, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058005831204354765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024930920801125465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014094258332625032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3260, + "weighted_flow_matching_loss": 0.01682747919112444, + "weighted_rkd_loss": 0.007047129166312516 + }, + { + "epoch": 5.589743589743589, + "flow_matching_loss": 0.01642184555530548, + "grad_norm": 0.17393237352371216, + "learning_rate": 8.802429710733745e-06, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00525958186481148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002003428607713431, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012522592302411794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3270, + "weighted_flow_matching_loss": 0.01642184555530548, + "weighted_rkd_loss": 0.006261296151205897 + }, + { + "epoch": 5.6068376068376065, + "flow_matching_loss": 0.015249202214181422, + "grad_norm": 0.15522201359272003, + "learning_rate": 8.71222165550757e-06, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004901192313991487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002037578250747174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011839963030070066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3280, + "weighted_flow_matching_loss": 0.015249202214181422, + "weighted_rkd_loss": 0.005919981515035033 + }, + { + "epoch": 5.6239316239316235, + "flow_matching_loss": 0.015745933819562196, + "grad_norm": 0.14974047243595123, + "learning_rate": 8.622288648075748e-06, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005023935274221003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020395138766616584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01208738461136818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3290, + "weighted_flow_matching_loss": 0.015745933819562196, + "weighted_rkd_loss": 0.00604369230568409 + }, + { + "epoch": 5.641025641025641, + "flow_matching_loss": 0.01888891002163291, + "grad_norm": 0.12535054981708527, + "learning_rate": 8.532634622399391e-06, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004384210170246661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001768164848908782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010536585142835975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3300, + "weighted_flow_matching_loss": 0.01888891002163291, + "weighted_rkd_loss": 0.0052682925714179875 + }, + { + "epoch": 5.6581196581196584, + "flow_matching_loss": 0.014870642125606537, + "grad_norm": 0.15545900166034698, + "learning_rate": 8.44326350023601e-06, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0048263138625770806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017976223956793548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011450250167399645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3310, + "weighted_flow_matching_loss": 0.014870642125606537, + "weighted_rkd_loss": 0.005725125083699822 + }, + { + "epoch": 5.6752136752136755, + "flow_matching_loss": 0.01560269109904766, + "grad_norm": 0.16620269417762756, + "learning_rate": 8.354179190968002e-06, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004752189433202147, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019955436349846424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011499922536313534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3320, + "weighted_flow_matching_loss": 0.01560269109904766, + "weighted_rkd_loss": 0.005749961268156767 + }, + { + "epoch": 5.6923076923076925, + "flow_matching_loss": 0.018268548417836428, + "grad_norm": 0.19080553948879242, + "learning_rate": 8.265385591431661e-06, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005082469270564616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020615262910723686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012226464878767728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3330, + "weighted_flow_matching_loss": 0.018268548417836428, + "weighted_rkd_loss": 0.006113232439383864 + }, + { + "epoch": 5.7094017094017095, + "flow_matching_loss": 0.016987710259854794, + "grad_norm": 0.176350936293602, + "learning_rate": 8.176886585746671e-06, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005133891617879272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020453140372410417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012313097342848777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3340, + "weighted_flow_matching_loss": 0.016987710259854794, + "weighted_rkd_loss": 0.006156548671424389 + }, + { + "epoch": 5.726495726495727, + "flow_matching_loss": 0.0147684495896101, + "grad_norm": 0.25896912813186646, + "learning_rate": 8.088686045146247e-06, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005305347871035338, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020323313889093696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012643027165904642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3350, + "weighted_flow_matching_loss": 0.0147684495896101, + "weighted_rkd_loss": 0.006321513582952321 + }, + { + "epoch": 5.743589743589744, + "flow_matching_loss": 0.016383587382733822, + "grad_norm": 0.16604885458946228, + "learning_rate": 8.00078782780775e-06, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005067420774139464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021367019740864634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01227154340595007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3360, + "weighted_flow_matching_loss": 0.016383587382733822, + "weighted_rkd_loss": 0.006135771702975035 + }, + { + "epoch": 5.760683760683761, + "flow_matching_loss": 0.018324275407940148, + "grad_norm": 0.22340555489063263, + "learning_rate": 7.913195778683979e-06, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005241550132632255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018835455761291088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012366645829752087, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3370, + "weighted_flow_matching_loss": 0.018324275407940148, + "weighted_rkd_loss": 0.006183322914876044 + }, + { + "epoch": 5.777777777777778, + "flow_matching_loss": 0.0162809194996953, + "grad_norm": 0.17706294357776642, + "learning_rate": 7.825913729334908e-06, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005191397480666637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020778504433110355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012460645381361246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3380, + "weighted_flow_matching_loss": 0.0162809194996953, + "weighted_rkd_loss": 0.006230322690680623 + }, + { + "epoch": 5.794871794871795, + "flow_matching_loss": 0.019178271014243364, + "grad_norm": 0.17093808948993683, + "learning_rate": 7.73894549776013e-06, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005111535452306271, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002210506598930806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012433577328920364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3390, + "weighted_flow_matching_loss": 0.019178271014243364, + "weighted_rkd_loss": 0.006216788664460182 + }, + { + "epoch": 5.811965811965812, + "flow_matching_loss": 0.015975886210799217, + "grad_norm": 0.13705652952194214, + "learning_rate": 7.65229488823184e-06, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006296888506039977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002588064025621861, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01518184095621109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3400, + "weighted_flow_matching_loss": 0.015975886210799217, + "weighted_rkd_loss": 0.007590920478105545 + }, + { + "epoch": 5.829059829059829, + "flow_matching_loss": 0.017226657923310996, + "grad_norm": 0.1626628041267395, + "learning_rate": 7.565965691128396e-06, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005771750584244728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024738962412811814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014017397444695235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3410, + "weighted_flow_matching_loss": 0.017226657923310996, + "weighted_rkd_loss": 0.007008698722347617 + }, + { + "epoch": 5.846153846153846, + "flow_matching_loss": 0.01667610052973032, + "grad_norm": 0.18348802626132965, + "learning_rate": 7.47996168276856e-06, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005051493598148227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018816259689629078, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011984613258391619, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3420, + "weighted_flow_matching_loss": 0.01667610052973032, + "weighted_rkd_loss": 0.005992306629195809 + }, + { + "epoch": 5.863247863247864, + "flow_matching_loss": 0.01668683849275112, + "grad_norm": 0.17090077698230743, + "learning_rate": 7.394286625246258e-06, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005260160844773054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023645029636099935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012884824723005294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3430, + "weighted_flow_matching_loss": 0.01668683849275112, + "weighted_rkd_loss": 0.006442412361502647 + }, + { + "epoch": 5.880341880341881, + "flow_matching_loss": 0.015944896265864372, + "grad_norm": 0.12532638013362885, + "learning_rate": 7.308944266266063e-06, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004895260930061341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019575991085730494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011748120933771134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3440, + "weighted_flow_matching_loss": 0.015944896265864372, + "weighted_rkd_loss": 0.005874060466885567 + }, + { + "epoch": 5.897435897435898, + "flow_matching_loss": 0.017557265516370536, + "grad_norm": 0.14915937185287476, + "learning_rate": 7.223938338979233e-06, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0045055771013721825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001900498615577817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010911652911454439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3450, + "weighted_flow_matching_loss": 0.017557265516370536, + "weighted_rkd_loss": 0.005455826455727219 + }, + { + "epoch": 5.914529914529915, + "flow_matching_loss": 0.018306993320584296, + "grad_norm": 0.15883104503154755, + "learning_rate": 7.139272561820397e-06, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004605731996707618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018143730005249381, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011025837063789368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3460, + "weighted_flow_matching_loss": 0.018306993320584296, + "weighted_rkd_loss": 0.005512918531894684 + }, + { + "epoch": 5.931623931623932, + "flow_matching_loss": 0.015725374035537244, + "grad_norm": 0.1372993290424347, + "learning_rate": 7.054950638344936e-06, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00528130738530308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002144762233365327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012707377038896083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3470, + "weighted_flow_matching_loss": 0.015725374035537244, + "weighted_rkd_loss": 0.006353688519448042 + }, + { + "epoch": 5.948717948717949, + "flow_matching_loss": 0.015682957507669924, + "grad_norm": 0.1382264494895935, + "learning_rate": 6.970976257066941e-06, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004367856332100928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018506806693039835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010586393298581243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3480, + "weighted_flow_matching_loss": 0.015682957507669924, + "weighted_rkd_loss": 0.005293196649290621 + }, + { + "epoch": 5.965811965811966, + "flow_matching_loss": 0.017090506479144098, + "grad_norm": 0.16394349932670593, + "learning_rate": 6.887353091297896e-06, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051696942886337634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020588471903465687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012398235965520144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3490, + "weighted_flow_matching_loss": 0.017090506479144098, + "weighted_rkd_loss": 0.006199117982760072 + }, + { + "epoch": 5.982905982905983, + "flow_matching_loss": 0.01643017726019025, + "grad_norm": 0.174118310213089, + "learning_rate": 6.804084798985965e-06, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004981178464367986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002098930941428989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012061287928372622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3500, + "weighted_flow_matching_loss": 0.01643017726019025, + "weighted_rkd_loss": 0.006030643964186311 + }, + { + "epoch": 6.0, + "flow_matching_loss": 0.01778987664729357, + "grad_norm": 0.193587064743042, + "learning_rate": 6.721175022556009e-06, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00433569997549057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017930015805177391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010464401543140411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3510, + "weighted_flow_matching_loss": 0.01778987664729357, + "weighted_rkd_loss": 0.0052322007715702055 + }, + { + "epoch": 6.017094017094017, + "flow_matching_loss": 0.01708282008767128, + "grad_norm": 0.16264356672763824, + "learning_rate": 6.638627388750245e-06, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004351393855176866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019219616777263581, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010624749260023237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3520, + "weighted_flow_matching_loss": 0.01708282008767128, + "weighted_rkd_loss": 0.005312374630011618 + }, + { + "epoch": 6.034188034188034, + "flow_matching_loss": 0.017007058300077916, + "grad_norm": 0.13111697137355804, + "learning_rate": 6.556445508469583e-06, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005663610715419054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002386491384822875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013713712897151708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3530, + "weighted_flow_matching_loss": 0.017007058300077916, + "weighted_rkd_loss": 0.006856856448575854 + }, + { + "epoch": 6.051282051282051, + "flow_matching_loss": 0.014417500142008067, + "grad_norm": 0.1420252025127411, + "learning_rate": 6.474632976615708e-06, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004468375444412231, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001813959248829633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010750709939748049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3540, + "weighted_flow_matching_loss": 0.014417500142008067, + "weighted_rkd_loss": 0.0053753549698740246 + }, + { + "epoch": 6.068376068376068, + "flow_matching_loss": 0.01604814138263464, + "grad_norm": 0.17824363708496094, + "learning_rate": 6.393193371933791e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004844561335630715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020488401991315187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01173796271905303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3550, + "weighted_flow_matching_loss": 0.01604814138263464, + "weighted_rkd_loss": 0.005868981359526515 + }, + { + "epoch": 6.085470085470085, + "flow_matching_loss": 0.01477200910449028, + "grad_norm": 0.17427536845207214, + "learning_rate": 6.312130256855977e-06, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044063873356208205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020096577471122147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010822432395070792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3560, + "weighted_flow_matching_loss": 0.01477200910449028, + "weighted_rkd_loss": 0.005411216197535396 + }, + { + "epoch": 6.102564102564102, + "flow_matching_loss": 0.0156622763723135, + "grad_norm": 0.13360509276390076, + "learning_rate": 6.23144717734552e-06, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004791768454015255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019348709029145539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011518407706171274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3570, + "weighted_flow_matching_loss": 0.0156622763723135, + "weighted_rkd_loss": 0.005759203853085637 + }, + { + "epoch": 6.119658119658119, + "flow_matching_loss": 0.016918635834008454, + "grad_norm": 0.15092743933200836, + "learning_rate": 6.151147662741695e-06, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004752983641810715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020754648372530936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011581432167440653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3580, + "weighted_flow_matching_loss": 0.016918635834008454, + "weighted_rkd_loss": 0.005790716083720326 + }, + { + "epoch": 6.136752136752137, + "flow_matching_loss": 0.016774132382124662, + "grad_norm": 0.13906005024909973, + "learning_rate": 6.071235225605415e-06, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004966419702395797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019617047160863875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011894544120877981, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3590, + "weighted_flow_matching_loss": 0.016774132382124662, + "weighted_rkd_loss": 0.005947272060438991 + }, + { + "epoch": 6.153846153846154, + "flow_matching_loss": 0.015961457509547472, + "grad_norm": 0.1445961594581604, + "learning_rate": 5.991713361565546e-06, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005250311200506985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002267113933339715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012767736334353685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3600, + "weighted_flow_matching_loss": 0.015961457509547472, + "weighted_rkd_loss": 0.006383868167176843 + }, + { + "epoch": 6.170940170940171, + "flow_matching_loss": 0.015231370739638806, + "grad_norm": 0.16246339678764343, + "learning_rate": 5.912585549166051e-06, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00495074198115617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018103821785189211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011711866315454244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3610, + "weighted_flow_matching_loss": 0.015231370739638806, + "weighted_rkd_loss": 0.005855933157727122 + }, + { + "epoch": 6.188034188034188, + "flow_matching_loss": 0.01628443393856287, + "grad_norm": 0.16113334894180298, + "learning_rate": 5.833855249713777e-06, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004770028358325362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018342768074944616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011374333407729864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3620, + "weighted_flow_matching_loss": 0.01628443393856287, + "weighted_rkd_loss": 0.005687166703864932 + }, + { + "epoch": 6.205128205128205, + "flow_matching_loss": 0.01632646331563592, + "grad_norm": 0.15477195382118225, + "learning_rate": 5.7555259071270645e-06, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004430649639107287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001690405874978751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010551705118268728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3630, + "weighted_flow_matching_loss": 0.01632646331563592, + "weighted_rkd_loss": 0.005275852559134364 + }, + { + "epoch": 6.222222222222222, + "flow_matching_loss": 0.014321347884833812, + "grad_norm": 0.16369390487670898, + "learning_rate": 5.677600947785117e-06, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005112451920285821, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002011882374063134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01223678607493639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3640, + "weighted_flow_matching_loss": 0.014321347884833812, + "weighted_rkd_loss": 0.006118393037468195 + }, + { + "epoch": 6.239316239316239, + "flow_matching_loss": 0.015911564882844686, + "grad_norm": 0.1967838704586029, + "learning_rate": 5.600083780378099e-06, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004792532348074019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018519766395911573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011437041219323874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3650, + "weighted_flow_matching_loss": 0.015911564882844686, + "weighted_rkd_loss": 0.005718520609661937 + }, + { + "epoch": 6.256410256410256, + "flow_matching_loss": 0.015266428980976344, + "grad_norm": 0.23702868819236755, + "learning_rate": 5.522977795758038e-06, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004600101965479553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017280104337260127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01092821448110044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3660, + "weighted_flow_matching_loss": 0.015266428980976344, + "weighted_rkd_loss": 0.00546410724055022 + }, + { + "epoch": 6.273504273504273, + "flow_matching_loss": 0.016469865664839746, + "grad_norm": 0.17626021802425385, + "learning_rate": 5.4462863667904805e-06, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005159393814392388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018100912333466113, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012128879036754369, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3670, + "weighted_flow_matching_loss": 0.016469865664839746, + "weighted_rkd_loss": 0.0060644395183771845 + }, + { + "epoch": 6.2905982905982905, + "flow_matching_loss": 0.01662844158709049, + "grad_norm": 0.2012549340724945, + "learning_rate": 5.370012848206983e-06, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004864910733886063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020002486533485355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011730070039629937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3680, + "weighted_flow_matching_loss": 0.01662844158709049, + "weighted_rkd_loss": 0.0058650350198149685 + }, + { + "epoch": 6.3076923076923075, + "flow_matching_loss": 0.015870072040706874, + "grad_norm": 0.14711031317710876, + "learning_rate": 5.29416057645834e-06, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004448941955342889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017441964941099287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010642080334946513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3690, + "weighted_flow_matching_loss": 0.015870072040706874, + "weighted_rkd_loss": 0.005321040167473257 + }, + { + "epoch": 6.3247863247863245, + "flow_matching_loss": 0.016839191410690547, + "grad_norm": 0.1470308154821396, + "learning_rate": 5.218732869568635e-06, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004482673481106758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00199831958161667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010963666439056396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3700, + "weighted_flow_matching_loss": 0.016839191410690547, + "weighted_rkd_loss": 0.005481833219528198 + }, + { + "epoch": 6.3418803418803416, + "flow_matching_loss": 0.015062647126615047, + "grad_norm": 0.15824107825756073, + "learning_rate": 5.14373302699013e-06, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005026099318638444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021568586002103985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012209057342261076, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3710, + "weighted_flow_matching_loss": 0.015062647126615047, + "weighted_rkd_loss": 0.006104528671130538 + }, + { + "epoch": 6.358974358974359, + "flow_matching_loss": 0.016755939926952124, + "grad_norm": 0.17414061725139618, + "learning_rate": 5.069164329458894e-06, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004771853005513549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017904155771248042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011334121599793434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3720, + "weighted_flow_matching_loss": 0.016755939926952124, + "weighted_rkd_loss": 0.005667060799896717 + }, + { + "epoch": 6.3760683760683765, + "flow_matching_loss": 0.017453379463404417, + "grad_norm": 0.18111607432365417, + "learning_rate": 4.995030038851342e-06, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004714675666764379, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018279231037013233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011257274448871613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3730, + "weighted_flow_matching_loss": 0.017453379463404417, + "weighted_rkd_loss": 0.0056286372244358064 + }, + { + "epoch": 6.3931623931623935, + "flow_matching_loss": 0.01427514124661684, + "grad_norm": 0.12923075258731842, + "learning_rate": 4.9213333980415e-06, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004160228231921792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016008496284484863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009921306138858199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3740, + "weighted_flow_matching_loss": 0.01427514124661684, + "weighted_rkd_loss": 0.004960653069429099 + }, + { + "epoch": 6.410256410256411, + "flow_matching_loss": 0.015912225097417833, + "grad_norm": 0.14404980838298798, + "learning_rate": 4.848077630759175e-06, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005185905285179615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022120023844763637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0125838128849864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3750, + "weighted_flow_matching_loss": 0.015912225097417833, + "weighted_rkd_loss": 0.0062919064424932 + }, + { + "epoch": 6.427350427350428, + "flow_matching_loss": 0.016386830247938633, + "grad_norm": 0.15200337767601013, + "learning_rate": 4.775265941448951e-06, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058306468650698665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022147785755805673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01387607241049409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3760, + "weighted_flow_matching_loss": 0.016386830247938633, + "weighted_rkd_loss": 0.006938036205247045 + }, + { + "epoch": 6.444444444444445, + "flow_matching_loss": 0.015980614442378284, + "grad_norm": 0.1789533942937851, + "learning_rate": 4.702901515129989e-06, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004540768498554826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002051534247584641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011133071221411227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3770, + "weighted_flow_matching_loss": 0.015980614442378284, + "weighted_rkd_loss": 0.005566535610705614 + }, + { + "epoch": 6.461538461538462, + "flow_matching_loss": 0.016609941609203815, + "grad_norm": 0.13359495997428894, + "learning_rate": 4.630987517256734e-06, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00556792204733938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021242573857307434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013260101620107888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3780, + "weighted_flow_matching_loss": 0.016609941609203815, + "weighted_rkd_loss": 0.006630050810053944 + }, + { + "epoch": 6.478632478632479, + "flow_matching_loss": 0.01568589322268963, + "grad_norm": 0.21237364411354065, + "learning_rate": 4.559527093580412e-06, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005020906263962388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020745873218402265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012116399919614195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3790, + "weighted_flow_matching_loss": 0.01568589322268963, + "weighted_rkd_loss": 0.006058199959807098 + }, + { + "epoch": 6.495726495726496, + "flow_matching_loss": 0.015495562460273504, + "grad_norm": 0.1483098566532135, + "learning_rate": 4.488523370011456e-06, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004437587270513177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001733652234543115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010608826857060194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3800, + "weighted_flow_matching_loss": 0.015495562460273504, + "weighted_rkd_loss": 0.005304413428530097 + }, + { + "epoch": 6.512820512820513, + "flow_matching_loss": 0.014806710742413997, + "grad_norm": 0.1273726522922516, + "learning_rate": 4.417979452482766e-06, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004741726303473115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019353451789356768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011418797774240375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3810, + "weighted_flow_matching_loss": 0.014806710742413997, + "weighted_rkd_loss": 0.005709398887120187 + }, + { + "epoch": 6.52991452991453, + "flow_matching_loss": 0.016286912932991983, + "grad_norm": 0.14928042888641357, + "learning_rate": 4.347898426813811e-06, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004060622188262641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015757541987113655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00969699863344431, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3820, + "weighted_flow_matching_loss": 0.016286912932991983, + "weighted_rkd_loss": 0.004848499316722155 + }, + { + "epoch": 6.547008547008547, + "flow_matching_loss": 0.01606764690950513, + "grad_norm": 0.15828359127044678, + "learning_rate": 4.278283358575694e-06, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00493182628415525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019138668780215085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011777519434690475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3830, + "weighted_flow_matching_loss": 0.01606764690950513, + "weighted_rkd_loss": 0.005888759717345237 + }, + { + "epoch": 6.564102564102564, + "flow_matching_loss": 0.01618655240163207, + "grad_norm": 0.22989870607852936, + "learning_rate": 4.2091372929570145e-06, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005280993832275272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002411555079743266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01297354269772768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3840, + "weighted_flow_matching_loss": 0.01618655240163207, + "weighted_rkd_loss": 0.00648677134886384 + }, + { + "epoch": 6.581196581196581, + "flow_matching_loss": 0.01933974390849471, + "grad_norm": 0.16607911884784698, + "learning_rate": 4.1404632546306934e-06, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00440069807227701, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018826108775101603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01068400703370571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3850, + "weighted_flow_matching_loss": 0.01933974390849471, + "weighted_rkd_loss": 0.005342003516852855 + }, + { + "epoch": 6.598290598290598, + "flow_matching_loss": 0.016087422892451287, + "grad_norm": 0.17650851607322693, + "learning_rate": 4.0722642476216286e-06, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004229920823127032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016205348074436188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01008037650026381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3860, + "weighted_flow_matching_loss": 0.016087422892451287, + "weighted_rkd_loss": 0.005040188250131905 + }, + { + "epoch": 6.615384615384615, + "flow_matching_loss": 0.015177225507795811, + "grad_norm": 0.1870707869529724, + "learning_rate": 4.00454325517533e-06, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005106892134062946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002013124746736139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012226908840239048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3870, + "weighted_flow_matching_loss": 0.015177225507795811, + "weighted_rkd_loss": 0.006113454420119524 + }, + { + "epoch": 6.632478632478632, + "flow_matching_loss": 0.015384182892739772, + "grad_norm": 0.165043905377388, + "learning_rate": 3.937303239627393e-06, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036144151352345945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001538443169556558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008767273370176553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3880, + "weighted_flow_matching_loss": 0.015384182892739772, + "weighted_rkd_loss": 0.004383636685088277 + }, + { + "epoch": 6.64957264957265, + "flow_matching_loss": 0.016996524203568696, + "grad_norm": 0.2328128218650818, + "learning_rate": 3.870547142273918e-06, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004557771165855229, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020668866171035917, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011182428896427154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3890, + "weighted_flow_matching_loss": 0.016996524203568696, + "weighted_rkd_loss": 0.005591214448213577 + }, + { + "epoch": 6.666666666666667, + "flow_matching_loss": 0.016605217289179564, + "grad_norm": 0.18834178149700165, + "learning_rate": 3.8042778832428736e-06, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004030448896810412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015587454079650342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009619643166661262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3900, + "weighted_flow_matching_loss": 0.016605217289179564, + "weighted_rkd_loss": 0.004809821583330631 + }, + { + "epoch": 6.683760683760684, + "flow_matching_loss": 0.019808900728821754, + "grad_norm": 0.17286179959774017, + "learning_rate": 3.7384983613663195e-06, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004534973669797182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016975721111521125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010767519474029541, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3910, + "weighted_flow_matching_loss": 0.019808900728821754, + "weighted_rkd_loss": 0.0053837597370147705 + }, + { + "epoch": 6.700854700854701, + "flow_matching_loss": 0.017337767593562602, + "grad_norm": 0.16282841563224792, + "learning_rate": 3.6732114540536503e-06, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041676078690215945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015985081496182828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009933723881840707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3920, + "weighted_flow_matching_loss": 0.017337767593562602, + "weighted_rkd_loss": 0.004966861940920353 + }, + { + "epoch": 6.717948717948718, + "flow_matching_loss": 0.016870119236409666, + "grad_norm": 0.15538223087787628, + "learning_rate": 3.608420017165683e-06, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005332878767512738, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020503941224887965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012716151541098952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3930, + "weighted_flow_matching_loss": 0.016870119236409666, + "weighted_rkd_loss": 0.006358075770549476 + }, + { + "epoch": 6.735042735042735, + "flow_matching_loss": 0.017999943252652882, + "grad_norm": 0.1980685442686081, + "learning_rate": 3.5441268848897664e-06, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00407553631812334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015552254626527429, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009706298168748618, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3940, + "weighted_flow_matching_loss": 0.017999943252652882, + "weighted_rkd_loss": 0.004853149084374309 + }, + { + "epoch": 6.752136752136752, + "flow_matching_loss": 0.014841801393777132, + "grad_norm": 0.14662159979343414, + "learning_rate": 3.480334869615792e-06, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043201320571824905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001651768758893013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010292032966390252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3950, + "weighted_flow_matching_loss": 0.014841801393777132, + "weighted_rkd_loss": 0.005146016483195126 + }, + { + "epoch": 6.769230769230769, + "flow_matching_loss": 0.014047346450388432, + "grad_norm": 0.1611618548631668, + "learning_rate": 3.417046761813155e-06, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004797124327160418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017810778925195336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011375326476991177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3960, + "weighted_flow_matching_loss": 0.014047346450388432, + "weighted_rkd_loss": 0.005687663238495589 + }, + { + "epoch": 6.786324786324786, + "flow_matching_loss": 0.016620884370058774, + "grad_norm": 0.2051396518945694, + "learning_rate": 3.3542653299087263e-06, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004709043470211327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018011134234257042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011219200305640697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3970, + "weighted_flow_matching_loss": 0.016620884370058774, + "weighted_rkd_loss": 0.0056096001528203486 + }, + { + "epoch": 6.803418803418803, + "flow_matching_loss": 0.013227695412933827, + "grad_norm": 0.1768273413181305, + "learning_rate": 3.2919933201657117e-06, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004239079589024186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019575843820348383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010435743583366274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3980, + "weighted_flow_matching_loss": 0.013227695412933827, + "weighted_rkd_loss": 0.005217871791683137 + }, + { + "epoch": 6.82051282051282, + "flow_matching_loss": 0.016534697357565166, + "grad_norm": 0.18357117474079132, + "learning_rate": 3.2302334565635584e-06, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004425921780057252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017305678804405034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010582411382347345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3990, + "weighted_flow_matching_loss": 0.016534697357565166, + "weighted_rkd_loss": 0.005291205691173672 + }, + { + "epoch": 6.837606837606837, + "flow_matching_loss": 0.014859104063361883, + "grad_norm": 0.21632759273052216, + "learning_rate": 3.1689884406787674e-06, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038600806845352055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016228868742473423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009343048324808478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4000, + "weighted_flow_matching_loss": 0.014859104063361883, + "weighted_rkd_loss": 0.004671524162404239 + }, + { + "epoch": 6.854700854700854, + "flow_matching_loss": 0.017187664564698935, + "grad_norm": 0.13118460774421692, + "learning_rate": 3.1082609515667424e-06, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004225519369356334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017153617111034692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010166400438174605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4010, + "weighted_flow_matching_loss": 0.017187664564698935, + "weighted_rkd_loss": 0.0050832002190873025 + }, + { + "epoch": 6.871794871794872, + "flow_matching_loss": 0.016087855119258164, + "grad_norm": 0.194352388381958, + "learning_rate": 3.048053645644594e-06, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004914767970331013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019320292049087584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011761565040796996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4020, + "weighted_flow_matching_loss": 0.016087855119258164, + "weighted_rkd_loss": 0.005880782520398498 + }, + { + "epoch": 6.888888888888889, + "flow_matching_loss": 0.015333417244255543, + "grad_norm": 0.15082167088985443, + "learning_rate": 2.988369156574916e-06, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003941808850504458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001569393661338836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009453011397272348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4030, + "weighted_flow_matching_loss": 0.015333417244255543, + "weighted_rkd_loss": 0.004726505698636174 + }, + { + "epoch": 6.905982905982906, + "flow_matching_loss": 0.016069279052317142, + "grad_norm": 0.13590292632579803, + "learning_rate": 2.9292100951506225e-06, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004573615314438939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018226096872240305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01096984026953578, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4040, + "weighted_flow_matching_loss": 0.016069279052317142, + "weighted_rkd_loss": 0.00548492013476789 + }, + { + "epoch": 6.923076923076923, + "flow_matching_loss": 0.015894505102187394, + "grad_norm": 0.16080982983112335, + "learning_rate": 2.8705790491807033e-06, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004575653979554772, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017367414198815822, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01088804923929274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4050, + "weighted_flow_matching_loss": 0.015894505102187394, + "weighted_rkd_loss": 0.00544402461964637 + }, + { + "epoch": 6.94017094017094, + "flow_matching_loss": 0.013970594946295023, + "grad_norm": 0.16656595468521118, + "learning_rate": 2.8124785833770433e-06, + "loss": 0.019, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004196502268314362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016817040625028313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010074708424508572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4060, + "weighted_flow_matching_loss": 0.013970594946295023, + "weighted_rkd_loss": 0.005037354212254286 + }, + { + "epoch": 6.957264957264957, + "flow_matching_loss": 0.017886763066053392, + "grad_norm": 0.15579462051391602, + "learning_rate": 2.754911239242241e-06, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004205319180618972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021213955304119738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010532033909112216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4070, + "weighted_flow_matching_loss": 0.017886763066053392, + "weighted_rkd_loss": 0.005266016954556108 + }, + { + "epoch": 6.9743589743589745, + "flow_matching_loss": 0.01466821664944291, + "grad_norm": 0.1741836667060852, + "learning_rate": 2.6978795349584204e-06, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0048158891731873155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001956576685188338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011588354967534542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4080, + "weighted_flow_matching_loss": 0.01466821664944291, + "weighted_rkd_loss": 0.005794177483767271 + }, + { + "epoch": 6.9914529914529915, + "flow_matching_loss": 0.016218194179236887, + "grad_norm": 0.15357689559459686, + "learning_rate": 2.641385965277088e-06, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004819849249906838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001966080185957253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011605778662487864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4090, + "weighted_flow_matching_loss": 0.016218194179236887, + "weighted_rkd_loss": 0.005802889331243932 + }, + { + "epoch": 7.0085470085470085, + "flow_matching_loss": 0.01698446348309517, + "grad_norm": 0.2071133255958557, + "learning_rate": 2.5854330014099923e-06, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004636326502077281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016064835479483008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01087913648225367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4100, + "weighted_flow_matching_loss": 0.01698446348309517, + "weighted_rkd_loss": 0.005439568241126835 + }, + { + "epoch": 7.0256410256410255, + "flow_matching_loss": 0.016615198738873006, + "grad_norm": 0.2097112238407135, + "learning_rate": 2.5300230909210298e-06, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004680479527451098, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001894661597907543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011255620792508126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4110, + "weighted_flow_matching_loss": 0.016615198738873006, + "weighted_rkd_loss": 0.005627810396254063 + }, + { + "epoch": 7.042735042735043, + "flow_matching_loss": 0.014118788111954928, + "grad_norm": 0.1714528203010559, + "learning_rate": 2.4751586576191983e-06, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004255960811860859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015853908960707485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010097312461584806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4120, + "weighted_flow_matching_loss": 0.014118788111954928, + "weighted_rkd_loss": 0.005048656230792403 + }, + { + "epoch": 7.05982905982906, + "flow_matching_loss": 0.01413761992007494, + "grad_norm": 0.14934183657169342, + "learning_rate": 2.420842101452538e-06, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004364838264882565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020897478796541693, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010819424223154784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4130, + "weighted_flow_matching_loss": 0.01413761992007494, + "weighted_rkd_loss": 0.005409712111577392 + }, + { + "epoch": 7.076923076923077, + "flow_matching_loss": 0.014027675800025462, + "grad_norm": 0.17164747416973114, + "learning_rate": 2.3670757984031816e-06, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037490071495994925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015432772459462285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009041291614994406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4140, + "weighted_flow_matching_loss": 0.014027675800025462, + "weighted_rkd_loss": 0.004520645807497203 + }, + { + "epoch": 7.094017094017094, + "flow_matching_loss": 0.013278034422546625, + "grad_norm": 0.12479699403047562, + "learning_rate": 2.313862100383404e-06, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036123374244198202, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014923112001270055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008716986002400517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4150, + "weighted_flow_matching_loss": 0.013278034422546625, + "weighted_rkd_loss": 0.004358493001200258 + }, + { + "epoch": 7.111111111111111, + "flow_matching_loss": 0.015609122812747955, + "grad_norm": 0.11537393182516098, + "learning_rate": 2.2612033351327406e-06, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0045773436082527045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002064816007623449, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011219503311440348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4160, + "weighted_flow_matching_loss": 0.015609122812747955, + "weighted_rkd_loss": 0.005609751655720174 + }, + { + "epoch": 7.128205128205128, + "flow_matching_loss": 0.014534916821867227, + "grad_norm": 0.1255650818347931, + "learning_rate": 2.2091018061161788e-06, + "loss": 0.0193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003910883702337742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016804726677946747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009502240177243947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4170, + "weighted_flow_matching_loss": 0.014534916821867227, + "weighted_rkd_loss": 0.004751120088621974 + }, + { + "epoch": 7.145299145299146, + "flow_matching_loss": 0.016385517176240684, + "grad_norm": 0.16010573506355286, + "learning_rate": 2.15755979242338e-06, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004504839330911637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001814916729927063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010824595438316464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4180, + "weighted_flow_matching_loss": 0.016385517176240684, + "weighted_rkd_loss": 0.005412297719158232 + }, + { + "epoch": 7.162393162393163, + "flow_matching_loss": 0.01727302558720112, + "grad_norm": 0.16287177801132202, + "learning_rate": 2.106579548669e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003930654772557318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013079404830932616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009169249981641769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4190, + "weighted_flow_matching_loss": 0.01727302558720112, + "weighted_rkd_loss": 0.004584624990820884 + }, + { + "epoch": 7.17948717948718, + "flow_matching_loss": 0.015191889088600873, + "grad_norm": 0.18671517074108124, + "learning_rate": 2.0561633048940477e-06, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004497651406563818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019035461591556668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010898848949000239, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4200, + "weighted_flow_matching_loss": 0.015191889088600873, + "weighted_rkd_loss": 0.005449424474500119 + }, + { + "epoch": 7.196581196581197, + "flow_matching_loss": 0.016087721660733224, + "grad_norm": 0.1463557481765747, + "learning_rate": 2.006313266468355e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004812970152124762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019313882337883114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011557328561320902, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4210, + "weighted_flow_matching_loss": 0.016087721660733224, + "weighted_rkd_loss": 0.005778664280660451 + }, + { + "epoch": 7.213675213675214, + "flow_matching_loss": 0.016287938598543404, + "grad_norm": 0.1706915646791458, + "learning_rate": 1.957031613994097e-06, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004160731332376599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016245927836280316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00994605547748506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4220, + "weighted_flow_matching_loss": 0.016287938598543404, + "weighted_rkd_loss": 0.00497302773874253 + }, + { + "epoch": 7.230769230769231, + "flow_matching_loss": 0.015654664393514393, + "grad_norm": 0.1440427303314209, + "learning_rate": 1.9083205032103995e-06, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004027574555948377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001447567099239677, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00950271631591022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4230, + "weighted_flow_matching_loss": 0.015654664393514393, + "weighted_rkd_loss": 0.00475135815795511 + }, + { + "epoch": 7.247863247863248, + "flow_matching_loss": 0.01606919979676604, + "grad_norm": 0.18371421098709106, + "learning_rate": 1.8601820648990536e-06, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004114628559909761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014564043085556477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009685661410912871, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4240, + "weighted_flow_matching_loss": 0.01606919979676604, + "weighted_rkd_loss": 0.004842830705456436 + }, + { + "epoch": 7.264957264957265, + "flow_matching_loss": 0.014553804695606232, + "grad_norm": 0.1450345814228058, + "learning_rate": 1.8126184047912936e-06, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038839376764371993, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014871912775561214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009255066746845841, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4250, + "weighted_flow_matching_loss": 0.014553804695606232, + "weighted_rkd_loss": 0.004627533373422921 + }, + { + "epoch": 7.282051282051282, + "flow_matching_loss": 0.013839024305343627, + "grad_norm": 0.22128815948963165, + "learning_rate": 1.7656316034757024e-06, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004872125829569996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002003580390010029, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0117478318978101, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4260, + "weighted_flow_matching_loss": 0.013839024305343627, + "weighted_rkd_loss": 0.00587391594890505 + }, + { + "epoch": 7.299145299145299, + "flow_matching_loss": 0.016494604852050544, + "grad_norm": 0.14832022786140442, + "learning_rate": 1.7192237163071816e-06, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003961826534941792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015516869083512575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00947534004226327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4270, + "weighted_flow_matching_loss": 0.016494604852050544, + "weighted_rkd_loss": 0.004737670021131635 + }, + { + "epoch": 7.316239316239316, + "flow_matching_loss": 0.015676535665988922, + "grad_norm": 0.14860793948173523, + "learning_rate": 1.6733967733170573e-06, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003999082953669131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002201343182241544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010199508955702185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4280, + "weighted_flow_matching_loss": 0.015676535665988922, + "weighted_rkd_loss": 0.005099754477851093 + }, + { + "epoch": 7.333333333333333, + "flow_matching_loss": 0.015473322197794914, + "grad_norm": 0.17381682991981506, + "learning_rate": 1.628152779124279e-06, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004379562707617879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016722492931876332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010431374609470367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4290, + "weighted_flow_matching_loss": 0.015473322197794914, + "weighted_rkd_loss": 0.005215687304735183 + }, + { + "epoch": 7.35042735042735, + "flow_matching_loss": 0.015086284559220075, + "grad_norm": 0.13543272018432617, + "learning_rate": 1.583493712847715e-06, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00412148276809603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015148779144510626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00975784338079393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4300, + "weighted_flow_matching_loss": 0.015086284559220075, + "weighted_rkd_loss": 0.004878921690396965 + }, + { + "epoch": 7.367521367521368, + "flow_matching_loss": 0.01674106726422906, + "grad_norm": 0.1808241754770279, + "learning_rate": 1.5394215280196084e-06, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003748697927221656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001673176372423768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009170572226867079, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4310, + "weighted_flow_matching_loss": 0.01674106726422906, + "weighted_rkd_loss": 0.0045852861134335395 + }, + { + "epoch": 7.384615384615385, + "flow_matching_loss": 0.01622048085555434, + "grad_norm": 0.24213089048862457, + "learning_rate": 1.4959381525000904e-06, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004719791142269969, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019229460158385336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011362528428435326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4320, + "weighted_flow_matching_loss": 0.01622048085555434, + "weighted_rkd_loss": 0.005681264214217663 + }, + { + "epoch": 7.401709401709402, + "flow_matching_loss": 0.01691405791789293, + "grad_norm": 0.136305570602417, + "learning_rate": 1.453045488392879e-06, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004566896031610668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017233178252354265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010857109911739826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4330, + "weighted_flow_matching_loss": 0.01691405791789293, + "weighted_rkd_loss": 0.005428554955869913 + }, + { + "epoch": 7.418803418803419, + "flow_matching_loss": 0.015587300807237626, + "grad_norm": 0.1919413059949875, + "learning_rate": 1.4107454119620484e-06, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004084520391188562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016211720067076385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009790212847292423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4340, + "weighted_flow_matching_loss": 0.015587300807237626, + "weighted_rkd_loss": 0.004895106423646212 + }, + { + "epoch": 7.435897435897436, + "flow_matching_loss": 0.016039892192929985, + "grad_norm": 0.18631604313850403, + "learning_rate": 1.3690397735499778e-06, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004030122491531074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018015784677118062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009861823450773955, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4350, + "weighted_flow_matching_loss": 0.016039892192929985, + "weighted_rkd_loss": 0.0049309117253869775 + }, + { + "epoch": 7.452991452991453, + "flow_matching_loss": 0.017094092350453137, + "grad_norm": 0.1710483431816101, + "learning_rate": 1.327930397496403e-06, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004542760131880641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001764396857470274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010849917074665427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4360, + "weighted_flow_matching_loss": 0.017094092350453137, + "weighted_rkd_loss": 0.005424958537332714 + }, + { + "epoch": 7.47008547008547, + "flow_matching_loss": 0.01391311539337039, + "grad_norm": 0.1865139752626419, + "learning_rate": 1.287419082058604e-06, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003823025431483984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001664593192981556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009310644119977951, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4370, + "weighted_flow_matching_loss": 0.01391311539337039, + "weighted_rkd_loss": 0.0046553220599889755 + }, + { + "epoch": 7.487179487179487, + "flow_matching_loss": 0.015736521501094104, + "grad_norm": 0.20026004314422607, + "learning_rate": 1.2475075993327589e-06, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004398590861819684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017495504696853458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010546732181683182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4380, + "weighted_flow_matching_loss": 0.015736521501094104, + "weighted_rkd_loss": 0.005273366090841591 + }, + { + "epoch": 7.504273504273504, + "flow_matching_loss": 0.015191830042749644, + "grad_norm": 0.12072700262069702, + "learning_rate": 1.208197695176413e-06, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004338941141031683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00170200303546153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01037988536991179, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4390, + "weighted_flow_matching_loss": 0.015191830042749644, + "weighted_rkd_loss": 0.005189942684955895 + }, + { + "epoch": 7.521367521367521, + "flow_matching_loss": 0.014817466028034687, + "grad_norm": 0.13009797036647797, + "learning_rate": 1.1694910891321259e-06, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003838592185638845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016448823153041303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009322066651657223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4400, + "weighted_flow_matching_loss": 0.014817466028034687, + "weighted_rkd_loss": 0.004661033325828612 + }, + { + "epoch": 7.538461538461538, + "flow_matching_loss": 0.017156030517071486, + "grad_norm": 0.18017660081386566, + "learning_rate": 1.1313894743522363e-06, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00444778697565198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019361308077350259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010831704782322048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4410, + "weighted_flow_matching_loss": 0.017156030517071486, + "weighted_rkd_loss": 0.005415852391161024 + }, + { + "epoch": 7.555555555555555, + "flow_matching_loss": 0.015430319868028165, + "grad_norm": 0.13498713076114655, + "learning_rate": 1.0938945175248016e-06, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003500660159625113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014626291289459915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00846394938416779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4420, + "weighted_flow_matching_loss": 0.015430319868028165, + "weighted_rkd_loss": 0.004231974692083895 + }, + { + "epoch": 7.572649572649572, + "flow_matching_loss": 0.016946085449308156, + "grad_norm": 0.14081688225269318, + "learning_rate": 1.0570078588007086e-06, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043074612505733965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018926878925412892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010507610347121953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4430, + "weighted_flow_matching_loss": 0.016946085449308156, + "weighted_rkd_loss": 0.005253805173560977 + }, + { + "epoch": 7.589743589743589, + "flow_matching_loss": 0.013496050518006087, + "grad_norm": 0.20047232508659363, + "learning_rate": 1.0207311117219009e-06, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003938631736673415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014799494645558297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00935721299611032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4440, + "weighted_flow_matching_loss": 0.013496050518006087, + "weighted_rkd_loss": 0.00467860649805516 + }, + { + "epoch": 7.6068376068376065, + "flow_matching_loss": 0.014154733717441558, + "grad_norm": 0.20217622816562653, + "learning_rate": 9.850658631508198e-07, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044426189735531805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018720550229772925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010757293039932846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4450, + "weighted_flow_matching_loss": 0.014154733717441558, + "weighted_rkd_loss": 0.005378646519966423 + }, + { + "epoch": 7.6239316239316235, + "flow_matching_loss": 0.01568802334368229, + "grad_norm": 0.13635766506195068, + "learning_rate": 9.500136732009751e-07, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004027505731210112, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016500390833243729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00970505066215992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4460, + "weighted_flow_matching_loss": 0.01568802334368229, + "weighted_rkd_loss": 0.00485252533107996 + }, + { + "epoch": 7.641025641025641, + "flow_matching_loss": 0.013338806852698327, + "grad_norm": 0.1933533251285553, + "learning_rate": 9.155760751687054e-07, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004383315378800034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017131514148786664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010479782242327929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4470, + "weighted_flow_matching_loss": 0.013338806852698327, + "weighted_rkd_loss": 0.005239891121163964 + }, + { + "epoch": 7.6581196581196584, + "flow_matching_loss": 0.015097386622801422, + "grad_norm": 0.1867772787809372, + "learning_rate": 8.817545754661155e-07, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004559310991317034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001779672270640731, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010898294346407055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4480, + "weighted_flow_matching_loss": 0.015097386622801422, + "weighted_rkd_loss": 0.005449147173203528 + }, + { + "epoch": 7.6752136752136755, + "flow_matching_loss": 0.017503133229911327, + "grad_norm": 0.1295897662639618, + "learning_rate": 8.485506535551668e-07, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039037192706018685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015860553365200758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009393493924289942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4490, + "weighted_flow_matching_loss": 0.017503133229911327, + "weighted_rkd_loss": 0.004696746962144971 + }, + { + "epoch": 7.6923076923076925, + "flow_matching_loss": 0.012718126643449068, + "grad_norm": 0.16545599699020386, + "learning_rate": 8.159657618829685e-07, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004243917670100928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016703173867426812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010158152924850583, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4500, + "weighted_flow_matching_loss": 0.012718126643449068, + "weighted_rkd_loss": 0.005079076462425292 + }, + { + "epoch": 7.7094017094017095, + "flow_matching_loss": 0.017403970565646888, + "grad_norm": 0.18544143438339233, + "learning_rate": 7.840013258182471e-07, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040193015011027455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016501099104061722, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009688712889328599, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4510, + "weighted_flow_matching_loss": 0.017403970565646888, + "weighted_rkd_loss": 0.0048443564446642995 + }, + { + "epoch": 7.726495726495727, + "flow_matching_loss": 0.017219954915344716, + "grad_norm": 0.13798153400421143, + "learning_rate": 7.526587435889809e-07, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035293982829898594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001458873733645305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00851767030544579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4520, + "weighted_flow_matching_loss": 0.017219954915344716, + "weighted_rkd_loss": 0.004258835152722895 + }, + { + "epoch": 7.743589743589744, + "flow_matching_loss": 0.01541247796267271, + "grad_norm": 0.139375701546669, + "learning_rate": 7.219393862212593e-07, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038150731241330506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014316341024823488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009061780339106918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4530, + "weighted_flow_matching_loss": 0.01541247796267271, + "weighted_rkd_loss": 0.004530890169553459 + }, + { + "epoch": 7.760683760683761, + "flow_matching_loss": 0.015388681180775165, + "grad_norm": 0.13138681650161743, + "learning_rate": 6.918445974792819e-07, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003366332477889955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001311596919549629, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008044261950999498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4540, + "weighted_flow_matching_loss": 0.015388681180775165, + "weighted_rkd_loss": 0.004022130975499749 + }, + { + "epoch": 7.777777777777778, + "flow_matching_loss": 0.014622988179326057, + "grad_norm": 0.11516177654266357, + "learning_rate": 6.6237569380661e-07, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038457690505310895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015454158827196806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009236953966319562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4550, + "weighted_flow_matching_loss": 0.014622988179326057, + "weighted_rkd_loss": 0.004618476983159781 + }, + { + "epoch": 7.794871794871795, + "flow_matching_loss": 0.014921532105654478, + "grad_norm": 0.13934694230556488, + "learning_rate": 6.335339642685495e-07, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037269066786393525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015626251930370926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00901643866673112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4560, + "weighted_flow_matching_loss": 0.014921532105654478, + "weighted_rkd_loss": 0.00450821933336556 + }, + { + "epoch": 7.811965811965812, + "flow_matching_loss": 0.015170859359204768, + "grad_norm": 0.20881541073322296, + "learning_rate": 6.05320670495787e-07, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00461851479485631, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017640399921219796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011001069517806172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4570, + "weighted_flow_matching_loss": 0.015170859359204768, + "weighted_rkd_loss": 0.005500534758903086 + }, + { + "epoch": 7.829059829059829, + "flow_matching_loss": 0.014338784758001566, + "grad_norm": 0.16537579894065857, + "learning_rate": 5.777370466291954e-07, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004272118653170764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001494649751111865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010038887057453394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4580, + "weighted_flow_matching_loss": 0.014338784758001566, + "weighted_rkd_loss": 0.005019443528726697 + }, + { + "epoch": 7.846153846153846, + "flow_matching_loss": 0.013995079416781664, + "grad_norm": 0.14324721693992615, + "learning_rate": 5.507842992658336e-07, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003364446433261037, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001459231018088758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008188123861327768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4590, + "weighted_flow_matching_loss": 0.013995079416781664, + "weighted_rkd_loss": 0.004094061930663884 + }, + { + "epoch": 7.863247863247864, + "flow_matching_loss": 0.0174401942640543, + "grad_norm": 0.1871991753578186, + "learning_rate": 5.24463607406192e-07, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00412391189020127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018717199098318815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010119543736800551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4600, + "weighted_flow_matching_loss": 0.0174401942640543, + "weighted_rkd_loss": 0.0050597718684002755 + }, + { + "epoch": 7.880341880341881, + "flow_matching_loss": 0.013484288938343525, + "grad_norm": 0.16377152502536774, + "learning_rate": 4.987761224025944e-07, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044992591720074415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016240559634752572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010622574388980866, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4610, + "weighted_flow_matching_loss": 0.013484288938343525, + "weighted_rkd_loss": 0.005311287194490433 + }, + { + "epoch": 7.897435897435898, + "flow_matching_loss": 0.017789938021451235, + "grad_norm": 0.1471613347530365, + "learning_rate": 4.737229679088556e-07, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004621709953062237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00174307432025671, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010986494272947312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4620, + "weighted_flow_matching_loss": 0.017789938021451235, + "weighted_rkd_loss": 0.005493247136473656 + }, + { + "epoch": 7.914529914529915, + "flow_matching_loss": 0.016959631443023683, + "grad_norm": 0.20595283806324005, + "learning_rate": 4.493052398311115e-07, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004121056059375405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017009781731758267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009943090286105872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4630, + "weighted_flow_matching_loss": 0.016959631443023683, + "weighted_rkd_loss": 0.004971545143052936 + }, + { + "epoch": 7.931623931623932, + "flow_matching_loss": 0.014979766122996807, + "grad_norm": 0.16615192592144012, + "learning_rate": 4.2552400627989043e-07, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004145107767544687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016354189720004797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009925634413957597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4640, + "weighted_flow_matching_loss": 0.014979766122996807, + "weighted_rkd_loss": 0.004962817206978798 + }, + { + "epoch": 7.948717948717949, + "flow_matching_loss": 0.015234688110649586, + "grad_norm": 0.14960739016532898, + "learning_rate": 4.023803075233923e-07, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004770138673484326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019104938895907254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011450771149247885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4650, + "weighted_flow_matching_loss": 0.015234688110649586, + "weighted_rkd_loss": 0.0057253855746239425 + }, + { + "epoch": 7.965811965811966, + "flow_matching_loss": 0.015797104593366383, + "grad_norm": 0.17209915816783905, + "learning_rate": 3.7987515594196997e-07, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044247575104236605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016803916427306832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010529906675219535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4660, + "weighted_flow_matching_loss": 0.015797104593366383, + "weighted_rkd_loss": 0.005264953337609768 + }, + { + "epoch": 7.982905982905983, + "flow_matching_loss": 0.015738171897828578, + "grad_norm": 0.13701865077018738, + "learning_rate": 3.580095359838648e-07, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004525296087376773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018712702090851962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010921862442046404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4670, + "weighted_flow_matching_loss": 0.015738171897828578, + "weighted_rkd_loss": 0.005460931221023202 + }, + { + "epoch": 8.0, + "flow_matching_loss": 0.015426522772759198, + "grad_norm": 0.14993874728679657, + "learning_rate": 3.3678440412212576e-07, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044903341913595796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018846069695428013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010865275282412767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4680, + "weighted_flow_matching_loss": 0.015426522772759198, + "weighted_rkd_loss": 0.005432637641206384 + }, + { + "epoch": 8.017094017094017, + "flow_matching_loss": 0.016985416971147062, + "grad_norm": 0.17564892768859863, + "learning_rate": 3.162006888127833e-07, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00391091569326818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014062735368497669, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00922810472548008, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4690, + "weighted_flow_matching_loss": 0.016985416971147062, + "weighted_rkd_loss": 0.00461405236274004 + }, + { + "epoch": 8.034188034188034, + "flow_matching_loss": 0.017536366637796164, + "grad_norm": 0.16656459867954254, + "learning_rate": 2.9625929045422327e-07, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039096885360777375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001401591202011332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009220968279987574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4700, + "weighted_flow_matching_loss": 0.017536366637796164, + "weighted_rkd_loss": 0.004610484139993787 + }, + { + "epoch": 8.051282051282051, + "flow_matching_loss": 0.016558579914271832, + "grad_norm": 0.16844052076339722, + "learning_rate": 2.7696108134781186e-07, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004956786707043648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017663251608610153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011679898528382183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4710, + "weighted_flow_matching_loss": 0.016558579914271832, + "weighted_rkd_loss": 0.005839949264191091 + }, + { + "epoch": 8.068376068376068, + "flow_matching_loss": 0.016213190834969284, + "grad_norm": 0.16253527998924255, + "learning_rate": 2.583069056597376e-07, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004380698874592781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020641373936086892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010825535003095865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4720, + "weighted_flow_matching_loss": 0.016213190834969284, + "weighted_rkd_loss": 0.005412767501547932 + }, + { + "epoch": 8.085470085470085, + "flow_matching_loss": 0.013580971863120794, + "grad_norm": 0.12660358846187592, + "learning_rate": 2.4029757938407116e-07, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004097833367995918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001680993870832026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009876660583540798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4730, + "weighted_flow_matching_loss": 0.013580971863120794, + "weighted_rkd_loss": 0.004938330291770399 + }, + { + "epoch": 8.102564102564102, + "flow_matching_loss": 0.01842335294932127, + "grad_norm": 0.18897368013858795, + "learning_rate": 2.229338903070871e-07, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004228601953946054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018422544933855533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010299458540976048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4740, + "weighted_flow_matching_loss": 0.01842335294932127, + "weighted_rkd_loss": 0.005149729270488024 + }, + { + "epoch": 8.11965811965812, + "flow_matching_loss": 0.01590350326150656, + "grad_norm": 0.20238161087036133, + "learning_rate": 2.0621659797279645e-07, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004119533277116716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015901223989203573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009829188929870725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4750, + "weighted_flow_matching_loss": 0.01590350326150656, + "weighted_rkd_loss": 0.004914594464935362 + }, + { + "epoch": 8.136752136752136, + "flow_matching_loss": 0.015774535667151212, + "grad_norm": 0.12638553977012634, + "learning_rate": 1.901464336497233e-07, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003907700185663998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001608267182018608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009423667518422007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4760, + "weighted_flow_matching_loss": 0.015774535667151212, + "weighted_rkd_loss": 0.004711833759211003 + }, + { + "epoch": 8.153846153846153, + "flow_matching_loss": 0.016548135597258807, + "grad_norm": 0.16260124742984772, + "learning_rate": 1.7472410029891527e-07, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036436608759686353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00162093035178259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00890825199894607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4770, + "weighted_flow_matching_loss": 0.016548135597258807, + "weighted_rkd_loss": 0.004454125999473035 + }, + { + "epoch": 8.17094017094017, + "flow_matching_loss": 0.015792224556207657, + "grad_norm": 0.1684548705816269, + "learning_rate": 1.5995027254319173e-07, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003674332913942635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001533985964488238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008882651710882783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4780, + "weighted_flow_matching_loss": 0.015792224556207657, + "weighted_rkd_loss": 0.004441325855441392 + }, + { + "epoch": 8.188034188034187, + "flow_matching_loss": 0.015294800698757171, + "grad_norm": 0.1381729245185852, + "learning_rate": 1.4582559663764372e-07, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004142222669906914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016778760007582604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009962321212515234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4790, + "weighted_flow_matching_loss": 0.015294800698757171, + "weighted_rkd_loss": 0.004981160606257617 + }, + { + "epoch": 8.205128205128204, + "flow_matching_loss": 0.016772223822772502, + "grad_norm": 0.161229208111763, + "learning_rate": 1.323506904413485e-07, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003599933069199324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016003339667804539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008800200140103697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4800, + "weighted_flow_matching_loss": 0.016772223822772502, + "weighted_rkd_loss": 0.0044001000700518485 + }, + { + "epoch": 8.222222222222221, + "flow_matching_loss": 0.014597597904503346, + "grad_norm": 0.15218256413936615, + "learning_rate": 1.195261433903594e-07, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004443337768316269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016796654323115944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010566340759396553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4810, + "weighted_flow_matching_loss": 0.014597597904503346, + "weighted_rkd_loss": 0.0052831703796982765 + }, + { + "epoch": 8.239316239316238, + "flow_matching_loss": 0.013690507505089045, + "grad_norm": 0.1815441995859146, + "learning_rate": 1.0735251647190969e-07, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004198707919567823, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016398903797380626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010037306183949112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4820, + "weighted_flow_matching_loss": 0.013690507505089045, + "weighted_rkd_loss": 0.005018653091974556 + }, + { + "epoch": 8.256410256410255, + "flow_matching_loss": 0.01400683168321848, + "grad_norm": 0.18531562387943268, + "learning_rate": 9.583034219987408e-08, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039456434780731795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015729700447991491, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009464256884530186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4830, + "weighted_flow_matching_loss": 0.01400683168321848, + "weighted_rkd_loss": 0.004732128442265093 + }, + { + "epoch": 8.273504273504274, + "flow_matching_loss": 0.013479845225811004, + "grad_norm": 0.12994004786014557, + "learning_rate": 8.496012459148717e-08, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00363312978297472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013098492461722344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008576108841225505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4840, + "weighted_flow_matching_loss": 0.013479845225811004, + "weighted_rkd_loss": 0.0042880544206127524 + }, + { + "epoch": 8.290598290598291, + "flow_matching_loss": 0.01754144886508584, + "grad_norm": 0.14384311437606812, + "learning_rate": 7.474233914527618e-08, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003767120442353189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014266954676713794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008960936358198524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4850, + "weighted_flow_matching_loss": 0.01754144886508584, + "weighted_rkd_loss": 0.004480468179099262 + }, + { + "epoch": 8.307692307692308, + "flow_matching_loss": 0.017515084147453307, + "grad_norm": 0.14004142582416534, + "learning_rate": 6.517743282028254e-08, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004615483921952546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017025555716827512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010933523252606391, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4860, + "weighted_flow_matching_loss": 0.017515084147453307, + "weighted_rkd_loss": 0.005466761626303196 + }, + { + "epoch": 8.324786324786325, + "flow_matching_loss": 0.01451965318992734, + "grad_norm": 0.18907830119132996, + "learning_rate": 5.6265824016492604e-08, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004115891247056425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019405572209507227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010172339901328088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4870, + "weighted_flow_matching_loss": 0.01451965318992734, + "weighted_rkd_loss": 0.005086169950664044 + }, + { + "epoch": 8.341880341880342, + "flow_matching_loss": 0.019001572486013175, + "grad_norm": 0.20190921425819397, + "learning_rate": 4.800790255654719e-08, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004398517566733063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019266181974671782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010723653389140964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4880, + "weighted_flow_matching_loss": 0.019001572486013175, + "weighted_rkd_loss": 0.005361826694570482 + }, + { + "epoch": 8.35897435897436, + "flow_matching_loss": 0.015850430447608234, + "grad_norm": 0.1569550633430481, + "learning_rate": 4.040402966867862e-08, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005174251366406679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001973703713156283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012322206469252705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4890, + "weighted_flow_matching_loss": 0.015850430447608234, + "weighted_rkd_loss": 0.0061611032346263524 + }, + { + "epoch": 8.376068376068377, + "flow_matching_loss": 0.01483481116592884, + "grad_norm": 0.15155529975891113, + "learning_rate": 3.3454537970925014e-08, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004053276544436812, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016709259594790637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009777478920295835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4900, + "weighted_flow_matching_loss": 0.01483481116592884, + "weighted_rkd_loss": 0.004888739460147918 + }, + { + "epoch": 8.393162393162394, + "flow_matching_loss": 0.018385430797934532, + "grad_norm": 0.17632077634334564, + "learning_rate": 2.715973145656192e-08, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003842594940215349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014750556205399335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009160245349630714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4910, + "weighted_flow_matching_loss": 0.018385430797934532, + "weighted_rkd_loss": 0.004580122674815357 + }, + { + "epoch": 8.41025641025641, + "flow_matching_loss": 0.014689198601990939, + "grad_norm": 0.1562715321779251, + "learning_rate": 2.151988548082129e-08, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004287013225257397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018384936323855073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010412520170211792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4920, + "weighted_flow_matching_loss": 0.014689198601990939, + "weighted_rkd_loss": 0.005206260085105896 + }, + { + "epoch": 8.427350427350428, + "flow_matching_loss": 0.01542881028726697, + "grad_norm": 0.17578746378421783, + "learning_rate": 1.6535246748837772e-08, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003949931729584933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001388546999078244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009288410376757383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4930, + "weighted_flow_matching_loss": 0.01542881028726697, + "weighted_rkd_loss": 0.004644205188378691 + }, + { + "epoch": 8.444444444444445, + "flow_matching_loss": 0.01600944958627224, + "grad_norm": 0.1673750877380371, + "learning_rate": 1.2206033304855702e-08, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038427982944995165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016363873146474361, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00932198385708034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4940, + "weighted_flow_matching_loss": 0.01600944958627224, + "weighted_rkd_loss": 0.00466099192854017 + }, + { + "epoch": 8.461538461538462, + "flow_matching_loss": 0.016356547176837922, + "grad_norm": 0.15482042729854584, + "learning_rate": 8.532434522695031e-09, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004332413501106203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001804306684061885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010469133732840419, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4950, + "weighted_flow_matching_loss": 0.016356547176837922, + "weighted_rkd_loss": 0.005234566866420209 + }, + { + "epoch": 8.478632478632479, + "flow_matching_loss": 0.01732535157352686, + "grad_norm": 0.16185925900936127, + "learning_rate": 5.5146110974679805e-09, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004459207039326429, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001728048489894718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010646462440490723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4960, + "weighted_flow_matching_loss": 0.01732535157352686, + "weighted_rkd_loss": 0.005323231220245361 + }, + { + "epoch": 8.495726495726496, + "flow_matching_loss": 0.015061072912067175, + "grad_norm": 0.23401165008544922, + "learning_rate": 3.1526950385496467e-09, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00431763231754303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017549702781252563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01039023487828672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4970, + "weighted_flow_matching_loss": 0.015061072912067175, + "weighted_rkd_loss": 0.00519511743914336 + }, + { + "epoch": 8.512820512820513, + "flow_matching_loss": 0.013777593802660703, + "grad_norm": 0.2139202058315277, + "learning_rate": 1.4467896637959666e-09, + "loss": 0.0193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00465741497464478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016942919231951238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011009121919050813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4980, + "weighted_flow_matching_loss": 0.013777593802660703, + "weighted_rkd_loss": 0.0055045609595254065 + }, + { + "epoch": 8.52991452991453, + "flow_matching_loss": 0.015110108815133572, + "grad_norm": 0.13214650750160217, + "learning_rate": 3.9696959503898733e-10, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003758502355776727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00138128298567608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00889828773215413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4990, + "weighted_flow_matching_loss": 0.015110108815133572, + "weighted_rkd_loss": 0.004449143866077065 + }, + { + "epoch": 8.547008547008547, + "flow_matching_loss": 0.014968675933778286, + "grad_norm": 0.17623348534107208, + "learning_rate": 3.280754806156416e-12, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004165530833415687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017792441823985428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01011030594818294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5000, + "weighted_flow_matching_loss": 0.014968675933778286, + "weighted_rkd_loss": 0.00505515297409147 + } + ], + "logging_steps": 10, + "max_steps": 5000, + "num_input_tokens_seen": 0, + "num_train_epochs": 9, + "save_steps": 5001, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": true + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 64, + "trial_name": null, + "trial_params": null +} diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/experiment_cfg/metadata.json b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..23dcfa871a88e56280f945357aef0296bd071711 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.778976917266846, + 0.33403685688972473, + 0.7223060727119446 + ], + "min": [ + -4.821934223175049, + -6.785110950469971, + 0.6994727253913879 + ], + "mean": [ + 2.577927505176127, + -2.1525618086425604, + 0.7003541659101092 + ], + "std": [ + 1.6720365948940972, + 1.3768414767482617, + 0.0009102921174316669 + ], + "q01": [ + -0.7605057497562686, + -5.518671947805291, + 0.6999764703640262 + ], + "q99": [ + 6.038879433964306, + -0.5168299981147954, + 0.7029182461269848 + ] + }, + "base_rotation": { + "max": [ + 0.0, + 0.0, + 1.0, + 1.0 + ], + "min": [ + 0.0, + 0.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0, + 0.0, + 0.24782371371751144, + 0.5971093380149053 + ], + "std": [ + 0.0, + 0.0, + 0.6642559499983499, + 0.3751891785184018 + ], + "q01": [ + 0.0, + 0.0, + -1.0, + 4.194193576416142e-06 + ], + "q99": [ + 0.0, + 0.0, + 1.0, + 1.0 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.8848381638526917, + 0.8569459915161133, + 1.053884744644165 + ], + "min": [ + -0.5828549861907959, + -0.8486236333847046, + -0.3446122705936432 + ], + "mean": [ + 0.25006971579930243, + -0.005716176032514759, + 0.4426027721228362 + ], + "std": [ + 0.15838384751435522, + 0.25141504139800513, + 0.20646816234620438 + ], + "q01": [ + -0.017003721077380903, + -0.47738320165276543, + 0.1917563409753314 + ], + "q99": [ + 0.5401029199137674, + 0.4706646609268599, + 0.6965586537529516 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999949932098389, + 0.9997354745864868, + 0.9900747537612915, + 0.9735639095306396 + ], + "min": [ + -0.9999914169311523, + -0.9999884366989136, + -0.9975369572639465, + 6.754255821306288e-08 + ], + "mean": [ + -0.22531628862537245, + 0.045669797538445966, + -0.06098055586474368, + 0.12292005493282424 + ], + "std": [ + 0.8506341437448487, + 0.2880763358322842, + 0.30816573781739026, + 0.16366165154486362 + ], + "q01": [ + -0.9957510593732032, + -0.5584741016891311, + -0.5371647381159943, + 0.0013784815838340492 + ], + "q99": [ + 0.9899375209218226, + 0.6327878884078038, + 0.5071539024531994, + 0.43555975763914345 + ] + }, + "gripper_qpos": { + "max": [ + 0.05289580300450325, + 0.00937709677964449 + ], + "min": [ + -0.011669117026031017, + -0.06307756155729294 + ], + "mean": [ + 0.03196506153371015, + -0.03229791582989581 + ], + "std": [ + 0.01171640503534049, + 0.01140016755425199 + ], + "q01": [ + 0.006290779996034005, + -0.04059717954551848 + ], + "q99": [ + 0.04038165838477291, + -0.006363076424932449 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.006112990930475949, + 0.002504955933349057, + -0.001639409958610569, + 0.0 + ], + "std": [ + 0.11478077265783658, + 0.1002662368589144, + 0.0807258815139021, + 0.0 + ], + "q01": [ + -0.16969970921613756, + -0.19919537649977875, + -0.15012139866371593, + 0.0 + ], + "q99": [ + 0.2466625453958827, + 0.25040842772679267, + 0.12777701094549537, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9253067479599785 + ], + "std": [ + 0.3792397014797737 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.1974093068624371 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + -0.006754222051360652, + 0.0027854853184013867, + -0.08087572130330477 + ], + "std": [ + 0.45779905420789546, + 0.40578975219522756, + 0.4062694098132682 + ], + "q01": [ + -0.968295960290036, + -0.9240402337368848, + -0.8249359802202183 + ], + "q99": [ + 0.8046634915851651, + 0.9212243025654695, + 0.838433711437593 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008633749251857347, + -0.020406457524911718, + -0.0027639446362883006 + ], + "std": [ + 0.11393972872407104, + 0.13300186268009506, + 0.12192288302920105 + ], + "q01": [ + -0.2668579353776187, + -0.40299251206771697, + -0.34712957892172586 + ], + "q99": [ + 0.32748281264383927, + 0.3179028969292779, + 0.33459405510046736 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3091635769453286 + ], + "std": [ + 0.9510122432876397 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.7643545615216344 + ] + } + } + }, + "modalities": { + "video": { + "robot0_eye_in_hand": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_left": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_right": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + } + }, + "state": { + "base_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "base_rotation": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "end_effector_position_relative": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation_relative": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "gripper_qpos": { + "absolute": true, + "rotation_type": null, + "shape": [ + 2 + ], + "continuous": true + } + }, + "action": { + "base_motion": { + "absolute": true, + "rotation_type": null, + "shape": [ + 4 + ], + "continuous": true + }, + "control_mode": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + }, + "end_effector_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation": { + "absolute": true, + "rotation_type": "axis_angle", + "shape": [ + 3 + ], + "continuous": true + }, + "gripper_close": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + } + } + }, + "embodiment_tag": "new_embodiment" + } +} \ No newline at end of file diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a3758b2dac9cab1274647a0a3e4647bb197ec7fc --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6f702e66e76c9c2a5773af5ec6b03a2aabb911ab03f04ce0b6ef1d2a8dca66a +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..c5fa6e6433c78dbcfdfe66fcc07909885778034b --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e043a1de49569a60b0c75a67e60a2f0aaa3fdaf767fab0d483b6a08343062969 +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..b2a1c3950609df867d30995f2b090d0416472d33 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e24fe3bebb040141f29e8a0118fe69ba061f79f1e93826f8fcc25d92f1efd45 +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..cefd7f914a530fda6608fc5aa9d81e6e53732142 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:58c5a5d3140c24a139d2991cadb8c07732e3165c7ecd098effda89e427b5bf1b +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..a7bb881e6a80800a024d5d0b9b433626b52276f4 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7ec5f38ed78266533eda4e31ed0a4fc31b0bdf4476a5791c5b52411e2397252c +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..46be4e3a7449fe6dddf7658364b8815ee6bd4760 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a6a03b57c8f4390db02278bf352d15420851e8a31f222227893b7860e66cdfb5 +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/rng_state.pth b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..ca778a0c5de38a9566a29f1aa51ae5bdbd5aaacd --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bb58fac01ea39bb458995cb653353e57b03a799c55f4f061c729e44186052c23 +size 14645 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/scheduler.pt b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..cddb85352b4d053450624108786800cbad81d6cb --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e19ead5779e3f8f2f666b7793de3a03d118f57f1b16a97eeb87bd732d18b445 +size 1465 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/model-00001-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/model-00001-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..8cc708d06de6fb76cb4bb7700c9f8db8ab7a4f3d --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/model-00001-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a13c7b0d9573493607c36f73c06ea8253f215c9f215a6990faf48ada8faa48ec +size 4999367032 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/model-00002-of-00002.safetensors b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/model-00002-of-00002.safetensors new file mode 100644 index 0000000000000000000000000000000000000000..46901d7a665fd069b59e6d6ce6d9584385346cd0 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/model-00002-of-00002.safetensors @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e77e954d8d63acce023c2920f9fa3a14cc17de8885b8470a789673ee5cfc8a3d +size 2664540348 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/rng_state.pth b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/rng_state.pth new file mode 100644 index 0000000000000000000000000000000000000000..896422b0e1d0a160b1edad8eb863f9f0d86e517d --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/rng_state.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d0de386febf3ad25d8b467c8d7ad77d32e3813fc3753e6c4855976f471fefa48 +size 14645 diff --git a/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/scheduler.pt b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..de4e41c4a6949df056e028b10a18e7f8cd7b7580 --- /dev/null +++ b/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:42d8e6d8e53e6b837084d0cc18fdd63ca9e40bd44a65d38eb52695647a8e84e8 +size 1465