Whalswp commited on
Commit
1ce4197
·
verified ·
1 Parent(s): b11e990

Add files using upload-large-folder tool

Browse files
Files changed (50) hide show
  1. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/config.json +101 -0
  2. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/experiment_cfg/metadata.json +431 -0
  3. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/model.safetensors.index.json +0 -0
  4. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/trainer_state.json +0 -0
  5. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/config.json +101 -0
  6. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/experiment_cfg/metadata.json +431 -0
  7. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/model.safetensors.index.json +0 -0
  8. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/trainer_state.json +0 -0
  9. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/config.json +101 -0
  10. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/experiment_cfg/metadata.json +431 -0
  11. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/model.safetensors.index.json +0 -0
  12. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/trainer_state.json +0 -0
  13. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/resolved_config.yaml +38 -0
  14. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/seed2.yaml +37 -0
  15. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/config.json +101 -0
  16. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/model.safetensors.index.json +0 -0
  17. FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/trainer_state.json +0 -0
  18. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/resolved_config.yaml +38 -0
  19. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/seed0.yaml +37 -0
  20. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/config.json +101 -0
  21. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json +431 -0
  22. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/model.safetensors.index.json +0 -0
  23. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/trainer_state.json +0 -0
  24. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/config.json +101 -0
  25. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json +431 -0
  26. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/model.safetensors.index.json +0 -0
  27. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/trainer_state.json +0 -0
  28. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/experiment_cfg/metadata.json +431 -0
  29. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/resolved_config.yaml +38 -0
  30. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/seed1.yaml +37 -0
  31. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/config.json +101 -0
  32. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json +431 -0
  33. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/model.safetensors.index.json +0 -0
  34. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/trainer_state.json +0 -0
  35. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/config.json +101 -0
  36. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json +431 -0
  37. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/model.safetensors.index.json +0 -0
  38. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/trainer_state.json +0 -0
  39. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/experiment_cfg/metadata.json +431 -0
  40. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/resolved_config.yaml +38 -0
  41. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/seed2.yaml +37 -0
  42. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/config.json +101 -0
  43. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json +431 -0
  44. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/model.safetensors.index.json +0 -0
  45. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/trainer_state.json +0 -0
  46. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/config.json +101 -0
  47. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json +431 -0
  48. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/model.safetensors.index.json +0 -0
  49. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/trainer_state.json +0 -0
  50. FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/experiment_cfg/metadata.json +431 -0
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/config.json ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_encoder_cls_ffn_hidden_dim": 2048,
63
+ "rkd_action_encoder_cls_num_heads": 8,
64
+ "rkd_action_encoder_cls_num_layers": 1,
65
+ "rkd_action_encoder_projector_dim": 512,
66
+ "rkd_action_encoder_projector_enabled": false,
67
+ "rkd_action_encoder_projector_pooling": "flatten",
68
+ "rkd_action_temp": 0.1,
69
+ "rkd_angle_loss_weight": 2.0,
70
+ "rkd_distance_loss_weight": 1.0,
71
+ "rkd_dtw_adaptive_floor": 0.05,
72
+ "rkd_dtw_adaptive_mix_lambda": 0.5,
73
+ "rkd_dtw_base_weight": 1.0,
74
+ "rkd_dtw_gripper_weight": 0.5,
75
+ "rkd_dtw_mix_mode": "distance_mix",
76
+ "rkd_dtw_normalize_components": true,
77
+ "rkd_dtw_pos_weight": 1.0,
78
+ "rkd_dtw_rot_weight": 1.0,
79
+ "rkd_dtw_timing_log_interval": 1000,
80
+ "rkd_enabled": true,
81
+ "rkd_exclude_diagonal": true,
82
+ "rkd_fm_loss_weight": 1.0,
83
+ "rkd_loss_type": "distance_angle",
84
+ "rkd_loss_weight": 0.5,
85
+ "rkd_loss_weight_end": 0.0,
86
+ "rkd_loss_weight_schedule": null,
87
+ "rkd_loss_weight_start": 0.1,
88
+ "rkd_raw_action_projector_dim": 512,
89
+ "rkd_raw_action_projector_enabled": false,
90
+ "rkd_raw_action_projector_pooling": "flatten",
91
+ "rkd_relation_mode": "token_pair_mean",
92
+ "rkd_student_reconstruction_head_dim": 512,
93
+ "rkd_student_reconstruction_head_enabled": true,
94
+ "rkd_student_reconstruction_head_hidden_dim": 512,
95
+ "rkd_student_reconstruction_head_pooling": "attention",
96
+ "rkd_teacher_source": "timewarp_action",
97
+ "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
98
+ "rkd_vlm_temp": 0.1,
99
+ "torch_dtype": "bfloat16",
100
+ "transformers_version": "4.51.3"
101
+ }
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.33403685688972473,
9
+ 0.7223060727119446
10
+ ],
11
+ "min": [
12
+ -4.821934223175049,
13
+ -6.785110950469971,
14
+ 0.6994727253913879
15
+ ],
16
+ "mean": [
17
+ 2.5886094736594543,
18
+ -2.1559664577316586,
19
+ 0.7003580854548973
20
+ ],
21
+ "std": [
22
+ 1.6717149343420803,
23
+ 1.3743695399555862,
24
+ 0.000940069542359506
25
+ ],
26
+ "q01": [
27
+ -0.7517815060249613,
28
+ -5.497849509089705,
29
+ 0.6999776914424374
30
+ ],
31
+ "q99": [
32
+ 6.060945690807028,
33
+ -0.5331499932493624,
34
+ 0.7028864260472257
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.24488939171811902,
54
+ 0.5968549204077488
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6659970734507764,
60
+ 0.37443298392262264
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 4.149924358336069e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8848381638526917,
78
+ 0.8569459915161133,
79
+ 1.053884744644165
80
+ ],
81
+ "min": [
82
+ -0.5828549861907959,
83
+ -0.8486236333847046,
84
+ -0.3446122705936432
85
+ ],
86
+ "mean": [
87
+ 0.25087486209693133,
88
+ -0.005738920462036393,
89
+ 0.43643948866361787
90
+ ],
91
+ "std": [
92
+ 0.15741544532518012,
93
+ 0.2547603106793105,
94
+ 0.2050434492495558
95
+ ],
96
+ "q01": [
97
+ -0.013990057288082561,
98
+ -0.4824599092623036,
99
+ 0.18825220353500022
100
+ ],
101
+ "q99": [
102
+ 0.5411989642841469,
103
+ 0.4754923657245658,
104
+ 0.6912323747882034
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999949932098389,
110
+ 0.9997354745864868,
111
+ 0.9900747537612915,
112
+ 0.9735639095306396
113
+ ],
114
+ "min": [
115
+ -0.9999914169311523,
116
+ -0.9999884366989136,
117
+ -0.9975369572639465,
118
+ 6.754255821306288e-08
119
+ ],
120
+ "mean": [
121
+ -0.22290331318740714,
122
+ 0.04579037037503981,
123
+ -0.05804094906527823,
124
+ 0.12347219024917105
125
+ ],
126
+ "std": [
127
+ 0.8519267931761344,
128
+ 0.2949394732270141,
129
+ 0.30026078832556913,
130
+ 0.16333086331829588
131
+ ],
132
+ "q01": [
133
+ -0.9957959450418518,
134
+ -0.5678869493126506,
135
+ -0.5318135475761875,
136
+ 0.0014160465686576505
137
+ ],
138
+ "q99": [
139
+ 0.9900692296898317,
140
+ 0.6357358195673684,
141
+ 0.5009847868015971,
142
+ 0.4287044482210845
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05289580300450325,
148
+ 0.00937709677964449
149
+ ],
150
+ "min": [
151
+ -0.011669117026031017,
152
+ -0.06307756155729294
153
+ ],
154
+ "mean": [
155
+ 0.03203268930378034,
156
+ -0.032331639271968075
157
+ ],
158
+ "std": [
159
+ 0.011607049173153697,
160
+ 0.011306286745872529
161
+ ],
162
+ "q01": [
163
+ 0.006429109393656672,
164
+ -0.04056586943958638
165
+ ],
166
+ "q99": [
167
+ 0.040377860142451545,
168
+ -0.006593889698673258
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.006050336852760097,
188
+ 0.002728916466628117,
189
+ -0.0017236811393598838,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11032167711449951,
194
+ 0.09824140301615687,
195
+ 0.07825208047515658,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.154721262094005,
200
+ -0.18955302988697728,
201
+ -0.14454065867754942,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.24492955247744197,
206
+ 0.2496797822190429,
207
+ 0.11966742325336034,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9294448715752568
220
+ ],
221
+ "std": [
222
+ 0.36897903503358326
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.21627562513311385
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.00826780165926883,
244
+ 0.003161997359379232,
245
+ -0.08157453651841007
246
+ ],
247
+ "std": [
248
+ 0.4545430944464769,
249
+ 0.41040151434986594,
250
+ 0.4102644251536544
251
+ ],
252
+ "q01": [
253
+ -0.9653185255999107,
254
+ -0.9255635199318754,
255
+ -0.8208437313341718
256
+ ],
257
+ "q99": [
258
+ 0.7987781084592441,
259
+ 0.9260844232330088,
260
+ 0.8357421812622201
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.007367024641295137,
276
+ -0.02060756343728361,
277
+ -0.001884833075998489
278
+ ],
279
+ "std": [
280
+ 0.1116768956941076,
281
+ 0.13249042272165212,
282
+ 0.12338908364725322
283
+ ],
284
+ "q01": [
285
+ -0.26894658663246845,
286
+ -0.39865607740339326,
287
+ -0.34393735191302655
288
+ ],
289
+ "q99": [
290
+ 0.3191674780171587,
291
+ 0.31133630397431267,
292
+ 0.33968476423051097
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3098994894722959
304
+ ],
305
+ "std": [
306
+ 0.9507718287498876
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7545620899936146
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-1000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/config.json ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_encoder_cls_ffn_hidden_dim": 2048,
63
+ "rkd_action_encoder_cls_num_heads": 8,
64
+ "rkd_action_encoder_cls_num_layers": 1,
65
+ "rkd_action_encoder_projector_dim": 512,
66
+ "rkd_action_encoder_projector_enabled": false,
67
+ "rkd_action_encoder_projector_pooling": "flatten",
68
+ "rkd_action_temp": 0.1,
69
+ "rkd_angle_loss_weight": 2.0,
70
+ "rkd_distance_loss_weight": 1.0,
71
+ "rkd_dtw_adaptive_floor": 0.05,
72
+ "rkd_dtw_adaptive_mix_lambda": 0.5,
73
+ "rkd_dtw_base_weight": 1.0,
74
+ "rkd_dtw_gripper_weight": 0.5,
75
+ "rkd_dtw_mix_mode": "distance_mix",
76
+ "rkd_dtw_normalize_components": true,
77
+ "rkd_dtw_pos_weight": 1.0,
78
+ "rkd_dtw_rot_weight": 1.0,
79
+ "rkd_dtw_timing_log_interval": 1000,
80
+ "rkd_enabled": true,
81
+ "rkd_exclude_diagonal": true,
82
+ "rkd_fm_loss_weight": 1.0,
83
+ "rkd_loss_type": "distance_angle",
84
+ "rkd_loss_weight": 0.5,
85
+ "rkd_loss_weight_end": 0.0,
86
+ "rkd_loss_weight_schedule": null,
87
+ "rkd_loss_weight_start": 0.1,
88
+ "rkd_raw_action_projector_dim": 512,
89
+ "rkd_raw_action_projector_enabled": false,
90
+ "rkd_raw_action_projector_pooling": "flatten",
91
+ "rkd_relation_mode": "token_pair_mean",
92
+ "rkd_student_reconstruction_head_dim": 512,
93
+ "rkd_student_reconstruction_head_enabled": true,
94
+ "rkd_student_reconstruction_head_hidden_dim": 512,
95
+ "rkd_student_reconstruction_head_pooling": "attention",
96
+ "rkd_teacher_source": "timewarp_action",
97
+ "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
98
+ "rkd_vlm_temp": 0.1,
99
+ "torch_dtype": "bfloat16",
100
+ "transformers_version": "4.51.3"
101
+ }
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.33403685688972473,
9
+ 0.7223060727119446
10
+ ],
11
+ "min": [
12
+ -4.821934223175049,
13
+ -6.785110950469971,
14
+ 0.6994727253913879
15
+ ],
16
+ "mean": [
17
+ 2.5886094736594543,
18
+ -2.1559664577316586,
19
+ 0.7003580854548973
20
+ ],
21
+ "std": [
22
+ 1.6717149343420803,
23
+ 1.3743695399555862,
24
+ 0.000940069542359506
25
+ ],
26
+ "q01": [
27
+ -0.7517815060249613,
28
+ -5.497849509089705,
29
+ 0.6999776914424374
30
+ ],
31
+ "q99": [
32
+ 6.060945690807028,
33
+ -0.5331499932493624,
34
+ 0.7028864260472257
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.24488939171811902,
54
+ 0.5968549204077488
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6659970734507764,
60
+ 0.37443298392262264
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 4.149924358336069e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8848381638526917,
78
+ 0.8569459915161133,
79
+ 1.053884744644165
80
+ ],
81
+ "min": [
82
+ -0.5828549861907959,
83
+ -0.8486236333847046,
84
+ -0.3446122705936432
85
+ ],
86
+ "mean": [
87
+ 0.25087486209693133,
88
+ -0.005738920462036393,
89
+ 0.43643948866361787
90
+ ],
91
+ "std": [
92
+ 0.15741544532518012,
93
+ 0.2547603106793105,
94
+ 0.2050434492495558
95
+ ],
96
+ "q01": [
97
+ -0.013990057288082561,
98
+ -0.4824599092623036,
99
+ 0.18825220353500022
100
+ ],
101
+ "q99": [
102
+ 0.5411989642841469,
103
+ 0.4754923657245658,
104
+ 0.6912323747882034
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999949932098389,
110
+ 0.9997354745864868,
111
+ 0.9900747537612915,
112
+ 0.9735639095306396
113
+ ],
114
+ "min": [
115
+ -0.9999914169311523,
116
+ -0.9999884366989136,
117
+ -0.9975369572639465,
118
+ 6.754255821306288e-08
119
+ ],
120
+ "mean": [
121
+ -0.22290331318740714,
122
+ 0.04579037037503981,
123
+ -0.05804094906527823,
124
+ 0.12347219024917105
125
+ ],
126
+ "std": [
127
+ 0.8519267931761344,
128
+ 0.2949394732270141,
129
+ 0.30026078832556913,
130
+ 0.16333086331829588
131
+ ],
132
+ "q01": [
133
+ -0.9957959450418518,
134
+ -0.5678869493126506,
135
+ -0.5318135475761875,
136
+ 0.0014160465686576505
137
+ ],
138
+ "q99": [
139
+ 0.9900692296898317,
140
+ 0.6357358195673684,
141
+ 0.5009847868015971,
142
+ 0.4287044482210845
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05289580300450325,
148
+ 0.00937709677964449
149
+ ],
150
+ "min": [
151
+ -0.011669117026031017,
152
+ -0.06307756155729294
153
+ ],
154
+ "mean": [
155
+ 0.03203268930378034,
156
+ -0.032331639271968075
157
+ ],
158
+ "std": [
159
+ 0.011607049173153697,
160
+ 0.011306286745872529
161
+ ],
162
+ "q01": [
163
+ 0.006429109393656672,
164
+ -0.04056586943958638
165
+ ],
166
+ "q99": [
167
+ 0.040377860142451545,
168
+ -0.006593889698673258
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.006050336852760097,
188
+ 0.002728916466628117,
189
+ -0.0017236811393598838,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11032167711449951,
194
+ 0.09824140301615687,
195
+ 0.07825208047515658,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.154721262094005,
200
+ -0.18955302988697728,
201
+ -0.14454065867754942,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.24492955247744197,
206
+ 0.2496797822190429,
207
+ 0.11966742325336034,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9294448715752568
220
+ ],
221
+ "std": [
222
+ 0.36897903503358326
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.21627562513311385
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.00826780165926883,
244
+ 0.003161997359379232,
245
+ -0.08157453651841007
246
+ ],
247
+ "std": [
248
+ 0.4545430944464769,
249
+ 0.41040151434986594,
250
+ 0.4102644251536544
251
+ ],
252
+ "q01": [
253
+ -0.9653185255999107,
254
+ -0.9255635199318754,
255
+ -0.8208437313341718
256
+ ],
257
+ "q99": [
258
+ 0.7987781084592441,
259
+ 0.9260844232330088,
260
+ 0.8357421812622201
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.007367024641295137,
276
+ -0.02060756343728361,
277
+ -0.001884833075998489
278
+ ],
279
+ "std": [
280
+ 0.1116768956941076,
281
+ 0.13249042272165212,
282
+ 0.12338908364725322
283
+ ],
284
+ "q01": [
285
+ -0.26894658663246845,
286
+ -0.39865607740339326,
287
+ -0.34393735191302655
288
+ ],
289
+ "q99": [
290
+ 0.3191674780171587,
291
+ 0.31133630397431267,
292
+ 0.33968476423051097
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3098994894722959
304
+ ],
305
+ "std": [
306
+ 0.9507718287498876
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7545620899936146
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-3000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/config.json ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_encoder_cls_ffn_hidden_dim": 2048,
63
+ "rkd_action_encoder_cls_num_heads": 8,
64
+ "rkd_action_encoder_cls_num_layers": 1,
65
+ "rkd_action_encoder_projector_dim": 512,
66
+ "rkd_action_encoder_projector_enabled": false,
67
+ "rkd_action_encoder_projector_pooling": "flatten",
68
+ "rkd_action_temp": 0.1,
69
+ "rkd_angle_loss_weight": 2.0,
70
+ "rkd_distance_loss_weight": 1.0,
71
+ "rkd_dtw_adaptive_floor": 0.05,
72
+ "rkd_dtw_adaptive_mix_lambda": 0.5,
73
+ "rkd_dtw_base_weight": 1.0,
74
+ "rkd_dtw_gripper_weight": 0.5,
75
+ "rkd_dtw_mix_mode": "distance_mix",
76
+ "rkd_dtw_normalize_components": true,
77
+ "rkd_dtw_pos_weight": 1.0,
78
+ "rkd_dtw_rot_weight": 1.0,
79
+ "rkd_dtw_timing_log_interval": 1000,
80
+ "rkd_enabled": true,
81
+ "rkd_exclude_diagonal": true,
82
+ "rkd_fm_loss_weight": 1.0,
83
+ "rkd_loss_type": "distance_angle",
84
+ "rkd_loss_weight": 0.5,
85
+ "rkd_loss_weight_end": 0.0,
86
+ "rkd_loss_weight_schedule": null,
87
+ "rkd_loss_weight_start": 0.1,
88
+ "rkd_raw_action_projector_dim": 512,
89
+ "rkd_raw_action_projector_enabled": false,
90
+ "rkd_raw_action_projector_pooling": "flatten",
91
+ "rkd_relation_mode": "token_pair_mean",
92
+ "rkd_student_reconstruction_head_dim": 512,
93
+ "rkd_student_reconstruction_head_enabled": true,
94
+ "rkd_student_reconstruction_head_hidden_dim": 512,
95
+ "rkd_student_reconstruction_head_pooling": "attention",
96
+ "rkd_teacher_source": "timewarp_action",
97
+ "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
98
+ "rkd_vlm_temp": 0.1,
99
+ "torch_dtype": "bfloat16",
100
+ "transformers_version": "4.51.3"
101
+ }
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.33403685688972473,
9
+ 0.7223060727119446
10
+ ],
11
+ "min": [
12
+ -4.821934223175049,
13
+ -6.785110950469971,
14
+ 0.6994727253913879
15
+ ],
16
+ "mean": [
17
+ 2.5886094736594543,
18
+ -2.1559664577316586,
19
+ 0.7003580854548973
20
+ ],
21
+ "std": [
22
+ 1.6717149343420803,
23
+ 1.3743695399555862,
24
+ 0.000940069542359506
25
+ ],
26
+ "q01": [
27
+ -0.7517815060249613,
28
+ -5.497849509089705,
29
+ 0.6999776914424374
30
+ ],
31
+ "q99": [
32
+ 6.060945690807028,
33
+ -0.5331499932493624,
34
+ 0.7028864260472257
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.24488939171811902,
54
+ 0.5968549204077488
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6659970734507764,
60
+ 0.37443298392262264
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 4.149924358336069e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8848381638526917,
78
+ 0.8569459915161133,
79
+ 1.053884744644165
80
+ ],
81
+ "min": [
82
+ -0.5828549861907959,
83
+ -0.8486236333847046,
84
+ -0.3446122705936432
85
+ ],
86
+ "mean": [
87
+ 0.25087486209693133,
88
+ -0.005738920462036393,
89
+ 0.43643948866361787
90
+ ],
91
+ "std": [
92
+ 0.15741544532518012,
93
+ 0.2547603106793105,
94
+ 0.2050434492495558
95
+ ],
96
+ "q01": [
97
+ -0.013990057288082561,
98
+ -0.4824599092623036,
99
+ 0.18825220353500022
100
+ ],
101
+ "q99": [
102
+ 0.5411989642841469,
103
+ 0.4754923657245658,
104
+ 0.6912323747882034
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999949932098389,
110
+ 0.9997354745864868,
111
+ 0.9900747537612915,
112
+ 0.9735639095306396
113
+ ],
114
+ "min": [
115
+ -0.9999914169311523,
116
+ -0.9999884366989136,
117
+ -0.9975369572639465,
118
+ 6.754255821306288e-08
119
+ ],
120
+ "mean": [
121
+ -0.22290331318740714,
122
+ 0.04579037037503981,
123
+ -0.05804094906527823,
124
+ 0.12347219024917105
125
+ ],
126
+ "std": [
127
+ 0.8519267931761344,
128
+ 0.2949394732270141,
129
+ 0.30026078832556913,
130
+ 0.16333086331829588
131
+ ],
132
+ "q01": [
133
+ -0.9957959450418518,
134
+ -0.5678869493126506,
135
+ -0.5318135475761875,
136
+ 0.0014160465686576505
137
+ ],
138
+ "q99": [
139
+ 0.9900692296898317,
140
+ 0.6357358195673684,
141
+ 0.5009847868015971,
142
+ 0.4287044482210845
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05289580300450325,
148
+ 0.00937709677964449
149
+ ],
150
+ "min": [
151
+ -0.011669117026031017,
152
+ -0.06307756155729294
153
+ ],
154
+ "mean": [
155
+ 0.03203268930378034,
156
+ -0.032331639271968075
157
+ ],
158
+ "std": [
159
+ 0.011607049173153697,
160
+ 0.011306286745872529
161
+ ],
162
+ "q01": [
163
+ 0.006429109393656672,
164
+ -0.04056586943958638
165
+ ],
166
+ "q99": [
167
+ 0.040377860142451545,
168
+ -0.006593889698673258
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.006050336852760097,
188
+ 0.002728916466628117,
189
+ -0.0017236811393598838,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11032167711449951,
194
+ 0.09824140301615687,
195
+ 0.07825208047515658,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.154721262094005,
200
+ -0.18955302988697728,
201
+ -0.14454065867754942,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.24492955247744197,
206
+ 0.2496797822190429,
207
+ 0.11966742325336034,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9294448715752568
220
+ ],
221
+ "std": [
222
+ 0.36897903503358326
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.21627562513311385
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.00826780165926883,
244
+ 0.003161997359379232,
245
+ -0.08157453651841007
246
+ ],
247
+ "std": [
248
+ 0.4545430944464769,
249
+ 0.41040151434986594,
250
+ 0.4102644251536544
251
+ ],
252
+ "q01": [
253
+ -0.9653185255999107,
254
+ -0.9255635199318754,
255
+ -0.8208437313341718
256
+ ],
257
+ "q99": [
258
+ 0.7987781084592441,
259
+ 0.9260844232330088,
260
+ 0.8357421812622201
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.007367024641295137,
276
+ -0.02060756343728361,
277
+ -0.001884833075998489
278
+ ],
279
+ "std": [
280
+ 0.1116768956941076,
281
+ 0.13249042272165212,
282
+ 0.12338908364725322
283
+ ],
284
+ "q01": [
285
+ -0.26894658663246845,
286
+ -0.39865607740339326,
287
+ -0.34393735191302655
288
+ ],
289
+ "q99": [
290
+ 0.3191674780171587,
291
+ 0.31133630397431267,
292
+ 0.33968476423051097
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3098994894722959
304
+ ],
305
+ "std": [
306
+ 0.9507718287498876
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7545620899936146
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed0/default/train_5shot_seed0_processingline_5k/checkpoint-5000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/resolved_config.yaml ADDED
@@ -0,0 +1,38 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: fewshot_39tasks_5shot_from_rkd_timewar_processingline_v2_seed2
2
+ policy_type: groot_RKDv3
3
+ base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2
5
+ dataset:
6
+ dataset_soup: fewshot_heldout_atomic39_human_5shot
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 64
10
+ seed: 42
11
+ filter_key_seed: 2
12
+ manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed2.json
13
+ save_total_limit: 4
14
+ model:
15
+ rkd_enabled: true
16
+ rkd_fm_loss_weight: 1.0
17
+ rkd_loss_weight: 0.5
18
+ rkd_loss_weight_schedule: null
19
+ phases:
20
+ - name: train_5shot_seed2_processingline_5k
21
+ max_steps: 5000
22
+ save_steps: 0
23
+ save_step_list:
24
+ - 1000
25
+ - 3000
26
+ - 5000
27
+ trainable:
28
+ preset: processing_line_only
29
+ tune_llm: false
30
+ tune_visual: false
31
+ tune_projector: false
32
+ tune_diffusion_model: false
33
+ losses:
34
+ rkd_enabled: true
35
+ rkd_fm_loss_weight: 1.0
36
+ rkd_loss_weight: 0.5
37
+ rkd_loss_weight_schedule: null
38
+ resolved_sweep: {}
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/seed2.yaml ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: fewshot_39tasks_5shot_from_rkd_timewar_processingline_v2_seed2
2
+ policy_type: groot_RKDv3
3
+ base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2
5
+ dataset:
6
+ dataset_soup: fewshot_heldout_atomic39_human_5shot
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 64
10
+ seed: 42
11
+ filter_key_seed: 2
12
+ manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed2.json
13
+ save_total_limit: 4
14
+ model:
15
+ rkd_enabled: true
16
+ rkd_fm_loss_weight: 1.0
17
+ rkd_loss_weight: 0.5
18
+ rkd_loss_weight_schedule: null
19
+ phases:
20
+ - name: train_5shot_seed2_processingline_5k
21
+ max_steps: 5000
22
+ save_steps: 0
23
+ save_step_list:
24
+ - 1000
25
+ - 3000
26
+ - 5000
27
+ trainable:
28
+ preset: processing_line_only
29
+ tune_llm: false
30
+ tune_visual: false
31
+ tune_projector: false
32
+ tune_diffusion_model: false
33
+ losses:
34
+ rkd_enabled: true
35
+ rkd_fm_loss_weight: 1.0
36
+ rkd_loss_weight: 0.5
37
+ rkd_loss_weight_schedule: null
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/config.json ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_encoder_cls_ffn_hidden_dim": 2048,
63
+ "rkd_action_encoder_cls_num_heads": 8,
64
+ "rkd_action_encoder_cls_num_layers": 1,
65
+ "rkd_action_encoder_projector_dim": 512,
66
+ "rkd_action_encoder_projector_enabled": false,
67
+ "rkd_action_encoder_projector_pooling": "flatten",
68
+ "rkd_action_temp": 0.1,
69
+ "rkd_angle_loss_weight": 2.0,
70
+ "rkd_distance_loss_weight": 1.0,
71
+ "rkd_dtw_adaptive_floor": 0.05,
72
+ "rkd_dtw_adaptive_mix_lambda": 0.5,
73
+ "rkd_dtw_base_weight": 1.0,
74
+ "rkd_dtw_gripper_weight": 0.5,
75
+ "rkd_dtw_mix_mode": "distance_mix",
76
+ "rkd_dtw_normalize_components": true,
77
+ "rkd_dtw_pos_weight": 1.0,
78
+ "rkd_dtw_rot_weight": 1.0,
79
+ "rkd_dtw_timing_log_interval": 1000,
80
+ "rkd_enabled": true,
81
+ "rkd_exclude_diagonal": true,
82
+ "rkd_fm_loss_weight": 1.0,
83
+ "rkd_loss_type": "distance_angle",
84
+ "rkd_loss_weight": 0.5,
85
+ "rkd_loss_weight_end": 0.0,
86
+ "rkd_loss_weight_schedule": null,
87
+ "rkd_loss_weight_start": 0.1,
88
+ "rkd_raw_action_projector_dim": 512,
89
+ "rkd_raw_action_projector_enabled": false,
90
+ "rkd_raw_action_projector_pooling": "flatten",
91
+ "rkd_relation_mode": "token_pair_mean",
92
+ "rkd_student_reconstruction_head_dim": 512,
93
+ "rkd_student_reconstruction_head_enabled": true,
94
+ "rkd_student_reconstruction_head_hidden_dim": 512,
95
+ "rkd_student_reconstruction_head_pooling": "attention",
96
+ "rkd_teacher_source": "timewarp_action",
97
+ "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
98
+ "rkd_vlm_temp": 0.1,
99
+ "torch_dtype": "bfloat16",
100
+ "transformers_version": "4.51.3"
101
+ }
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewar_Processingline_v2/seed2/default/train_5shot_seed2_processingline_5k/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/resolved_config.yaml ADDED
@@ -0,0 +1,38 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: fewshot_39tasks_5shot_from_rkd_timewarp_fullactionhead_seed0
2
+ policy_type: groot_RKDv3
3
+ base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0
5
+ dataset:
6
+ dataset_soup: fewshot_heldout_atomic39_human_5shot
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 64
10
+ seed: 42
11
+ filter_key_seed: 0
12
+ manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed0.json
13
+ save_total_limit: 4
14
+ model:
15
+ rkd_enabled: true
16
+ rkd_fm_loss_weight: 1.0
17
+ rkd_loss_weight: 0.5
18
+ rkd_loss_weight_schedule: null
19
+ phases:
20
+ - name: train_5shot_seed0_fullactionhead_5k
21
+ max_steps: 5000
22
+ save_steps: 0
23
+ save_step_list:
24
+ - 1000
25
+ - 3000
26
+ - 5000
27
+ save_final_model: false
28
+ trainable:
29
+ tune_llm: false
30
+ tune_visual: false
31
+ tune_projector: true
32
+ tune_diffusion_model: true
33
+ losses:
34
+ rkd_enabled: true
35
+ rkd_fm_loss_weight: 1.0
36
+ rkd_loss_weight: 0.5
37
+ rkd_loss_weight_schedule: null
38
+ resolved_sweep: {}
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/seed0.yaml ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: fewshot_39tasks_5shot_from_rkd_timewarp_fullactionhead_seed0
2
+ policy_type: groot_RKDv3
3
+ base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0
5
+ dataset:
6
+ dataset_soup: fewshot_heldout_atomic39_human_5shot
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 64
10
+ seed: 42
11
+ filter_key_seed: 0
12
+ manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed0.json
13
+ save_total_limit: 4
14
+ model:
15
+ rkd_enabled: true
16
+ rkd_fm_loss_weight: 1.0
17
+ rkd_loss_weight: 0.5
18
+ rkd_loss_weight_schedule: null
19
+ phases:
20
+ - name: train_5shot_seed0_fullactionhead_5k
21
+ max_steps: 5000
22
+ save_steps: 0
23
+ save_step_list:
24
+ - 1000
25
+ - 3000
26
+ - 5000
27
+ save_final_model: false
28
+ trainable:
29
+ tune_llm: false
30
+ tune_visual: false
31
+ tune_projector: true
32
+ tune_diffusion_model: true
33
+ losses:
34
+ rkd_enabled: true
35
+ rkd_fm_loss_weight: 1.0
36
+ rkd_loss_weight: 0.5
37
+ rkd_loss_weight_schedule: null
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/config.json ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_encoder_cls_ffn_hidden_dim": 2048,
63
+ "rkd_action_encoder_cls_num_heads": 8,
64
+ "rkd_action_encoder_cls_num_layers": 1,
65
+ "rkd_action_encoder_projector_dim": 512,
66
+ "rkd_action_encoder_projector_enabled": false,
67
+ "rkd_action_encoder_projector_pooling": "flatten",
68
+ "rkd_action_temp": 0.1,
69
+ "rkd_angle_loss_weight": 2.0,
70
+ "rkd_distance_loss_weight": 1.0,
71
+ "rkd_dtw_adaptive_floor": 0.05,
72
+ "rkd_dtw_adaptive_mix_lambda": 0.5,
73
+ "rkd_dtw_base_weight": 1.0,
74
+ "rkd_dtw_gripper_weight": 0.5,
75
+ "rkd_dtw_mix_mode": "distance_mix",
76
+ "rkd_dtw_normalize_components": true,
77
+ "rkd_dtw_pos_weight": 1.0,
78
+ "rkd_dtw_rot_weight": 1.0,
79
+ "rkd_dtw_timing_log_interval": 1000,
80
+ "rkd_enabled": true,
81
+ "rkd_exclude_diagonal": true,
82
+ "rkd_fm_loss_weight": 1.0,
83
+ "rkd_loss_type": "distance_angle",
84
+ "rkd_loss_weight": 0.5,
85
+ "rkd_loss_weight_end": 0.0,
86
+ "rkd_loss_weight_schedule": null,
87
+ "rkd_loss_weight_start": 0.1,
88
+ "rkd_raw_action_projector_dim": 512,
89
+ "rkd_raw_action_projector_enabled": false,
90
+ "rkd_raw_action_projector_pooling": "flatten",
91
+ "rkd_relation_mode": "token_pair_mean",
92
+ "rkd_student_reconstruction_head_dim": 512,
93
+ "rkd_student_reconstruction_head_enabled": true,
94
+ "rkd_student_reconstruction_head_hidden_dim": 512,
95
+ "rkd_student_reconstruction_head_pooling": "attention",
96
+ "rkd_teacher_source": "timewarp_action",
97
+ "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
98
+ "rkd_vlm_temp": 0.1,
99
+ "torch_dtype": "bfloat16",
100
+ "transformers_version": "4.51.3"
101
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.33403685688972473,
9
+ 0.7223060727119446
10
+ ],
11
+ "min": [
12
+ -4.821934223175049,
13
+ -6.785110950469971,
14
+ 0.6994727253913879
15
+ ],
16
+ "mean": [
17
+ 2.5886094736594543,
18
+ -2.1559664577316586,
19
+ 0.7003580854548973
20
+ ],
21
+ "std": [
22
+ 1.6717149343420803,
23
+ 1.3743695399555862,
24
+ 0.000940069542359506
25
+ ],
26
+ "q01": [
27
+ -0.7517815060249613,
28
+ -5.497849509089705,
29
+ 0.6999776914424374
30
+ ],
31
+ "q99": [
32
+ 6.060945690807028,
33
+ -0.5331499932493624,
34
+ 0.7028864260472257
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.24488939171811902,
54
+ 0.5968549204077488
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6659970734507764,
60
+ 0.37443298392262264
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 4.149924358336069e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8848381638526917,
78
+ 0.8569459915161133,
79
+ 1.053884744644165
80
+ ],
81
+ "min": [
82
+ -0.5828549861907959,
83
+ -0.8486236333847046,
84
+ -0.3446122705936432
85
+ ],
86
+ "mean": [
87
+ 0.25087486209693133,
88
+ -0.005738920462036393,
89
+ 0.43643948866361787
90
+ ],
91
+ "std": [
92
+ 0.15741544532518012,
93
+ 0.2547603106793105,
94
+ 0.2050434492495558
95
+ ],
96
+ "q01": [
97
+ -0.013990057288082561,
98
+ -0.4824599092623036,
99
+ 0.18825220353500022
100
+ ],
101
+ "q99": [
102
+ 0.5411989642841469,
103
+ 0.4754923657245658,
104
+ 0.6912323747882034
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999949932098389,
110
+ 0.9997354745864868,
111
+ 0.9900747537612915,
112
+ 0.9735639095306396
113
+ ],
114
+ "min": [
115
+ -0.9999914169311523,
116
+ -0.9999884366989136,
117
+ -0.9975369572639465,
118
+ 6.754255821306288e-08
119
+ ],
120
+ "mean": [
121
+ -0.22290331318740714,
122
+ 0.04579037037503981,
123
+ -0.05804094906527823,
124
+ 0.12347219024917105
125
+ ],
126
+ "std": [
127
+ 0.8519267931761344,
128
+ 0.2949394732270141,
129
+ 0.30026078832556913,
130
+ 0.16333086331829588
131
+ ],
132
+ "q01": [
133
+ -0.9957959450418518,
134
+ -0.5678869493126506,
135
+ -0.5318135475761875,
136
+ 0.0014160465686576505
137
+ ],
138
+ "q99": [
139
+ 0.9900692296898317,
140
+ 0.6357358195673684,
141
+ 0.5009847868015971,
142
+ 0.4287044482210845
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05289580300450325,
148
+ 0.00937709677964449
149
+ ],
150
+ "min": [
151
+ -0.011669117026031017,
152
+ -0.06307756155729294
153
+ ],
154
+ "mean": [
155
+ 0.03203268930378034,
156
+ -0.032331639271968075
157
+ ],
158
+ "std": [
159
+ 0.011607049173153697,
160
+ 0.011306286745872529
161
+ ],
162
+ "q01": [
163
+ 0.006429109393656672,
164
+ -0.04056586943958638
165
+ ],
166
+ "q99": [
167
+ 0.040377860142451545,
168
+ -0.006593889698673258
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.006050336852760097,
188
+ 0.002728916466628117,
189
+ -0.0017236811393598838,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11032167711449951,
194
+ 0.09824140301615687,
195
+ 0.07825208047515658,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.154721262094005,
200
+ -0.18955302988697728,
201
+ -0.14454065867754942,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.24492955247744197,
206
+ 0.2496797822190429,
207
+ 0.11966742325336034,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9294448715752568
220
+ ],
221
+ "std": [
222
+ 0.36897903503358326
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.21627562513311385
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.00826780165926883,
244
+ 0.003161997359379232,
245
+ -0.08157453651841007
246
+ ],
247
+ "std": [
248
+ 0.4545430944464769,
249
+ 0.41040151434986594,
250
+ 0.4102644251536544
251
+ ],
252
+ "q01": [
253
+ -0.9653185255999107,
254
+ -0.9255635199318754,
255
+ -0.8208437313341718
256
+ ],
257
+ "q99": [
258
+ 0.7987781084592441,
259
+ 0.9260844232330088,
260
+ 0.8357421812622201
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.007367024641295137,
276
+ -0.02060756343728361,
277
+ -0.001884833075998489
278
+ ],
279
+ "std": [
280
+ 0.1116768956941076,
281
+ 0.13249042272165212,
282
+ 0.12338908364725322
283
+ ],
284
+ "q01": [
285
+ -0.26894658663246845,
286
+ -0.39865607740339326,
287
+ -0.34393735191302655
288
+ ],
289
+ "q99": [
290
+ 0.3191674780171587,
291
+ 0.31133630397431267,
292
+ 0.33968476423051097
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3098994894722959
304
+ ],
305
+ "std": [
306
+ 0.9507718287498876
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7545620899936146
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-1000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/config.json ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_encoder_cls_ffn_hidden_dim": 2048,
63
+ "rkd_action_encoder_cls_num_heads": 8,
64
+ "rkd_action_encoder_cls_num_layers": 1,
65
+ "rkd_action_encoder_projector_dim": 512,
66
+ "rkd_action_encoder_projector_enabled": false,
67
+ "rkd_action_encoder_projector_pooling": "flatten",
68
+ "rkd_action_temp": 0.1,
69
+ "rkd_angle_loss_weight": 2.0,
70
+ "rkd_distance_loss_weight": 1.0,
71
+ "rkd_dtw_adaptive_floor": 0.05,
72
+ "rkd_dtw_adaptive_mix_lambda": 0.5,
73
+ "rkd_dtw_base_weight": 1.0,
74
+ "rkd_dtw_gripper_weight": 0.5,
75
+ "rkd_dtw_mix_mode": "distance_mix",
76
+ "rkd_dtw_normalize_components": true,
77
+ "rkd_dtw_pos_weight": 1.0,
78
+ "rkd_dtw_rot_weight": 1.0,
79
+ "rkd_dtw_timing_log_interval": 1000,
80
+ "rkd_enabled": true,
81
+ "rkd_exclude_diagonal": true,
82
+ "rkd_fm_loss_weight": 1.0,
83
+ "rkd_loss_type": "distance_angle",
84
+ "rkd_loss_weight": 0.5,
85
+ "rkd_loss_weight_end": 0.0,
86
+ "rkd_loss_weight_schedule": null,
87
+ "rkd_loss_weight_start": 0.1,
88
+ "rkd_raw_action_projector_dim": 512,
89
+ "rkd_raw_action_projector_enabled": false,
90
+ "rkd_raw_action_projector_pooling": "flatten",
91
+ "rkd_relation_mode": "token_pair_mean",
92
+ "rkd_student_reconstruction_head_dim": 512,
93
+ "rkd_student_reconstruction_head_enabled": true,
94
+ "rkd_student_reconstruction_head_hidden_dim": 512,
95
+ "rkd_student_reconstruction_head_pooling": "attention",
96
+ "rkd_teacher_source": "timewarp_action",
97
+ "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
98
+ "rkd_vlm_temp": 0.1,
99
+ "torch_dtype": "bfloat16",
100
+ "transformers_version": "4.51.3"
101
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.33403685688972473,
9
+ 0.7223060727119446
10
+ ],
11
+ "min": [
12
+ -4.821934223175049,
13
+ -6.785110950469971,
14
+ 0.6994727253913879
15
+ ],
16
+ "mean": [
17
+ 2.5886094736594543,
18
+ -2.1559664577316586,
19
+ 0.7003580854548973
20
+ ],
21
+ "std": [
22
+ 1.6717149343420803,
23
+ 1.3743695399555862,
24
+ 0.000940069542359506
25
+ ],
26
+ "q01": [
27
+ -0.7517815060249613,
28
+ -5.497849509089705,
29
+ 0.6999776914424374
30
+ ],
31
+ "q99": [
32
+ 6.060945690807028,
33
+ -0.5331499932493624,
34
+ 0.7028864260472257
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.24488939171811902,
54
+ 0.5968549204077488
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6659970734507764,
60
+ 0.37443298392262264
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 4.149924358336069e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8848381638526917,
78
+ 0.8569459915161133,
79
+ 1.053884744644165
80
+ ],
81
+ "min": [
82
+ -0.5828549861907959,
83
+ -0.8486236333847046,
84
+ -0.3446122705936432
85
+ ],
86
+ "mean": [
87
+ 0.25087486209693133,
88
+ -0.005738920462036393,
89
+ 0.43643948866361787
90
+ ],
91
+ "std": [
92
+ 0.15741544532518012,
93
+ 0.2547603106793105,
94
+ 0.2050434492495558
95
+ ],
96
+ "q01": [
97
+ -0.013990057288082561,
98
+ -0.4824599092623036,
99
+ 0.18825220353500022
100
+ ],
101
+ "q99": [
102
+ 0.5411989642841469,
103
+ 0.4754923657245658,
104
+ 0.6912323747882034
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999949932098389,
110
+ 0.9997354745864868,
111
+ 0.9900747537612915,
112
+ 0.9735639095306396
113
+ ],
114
+ "min": [
115
+ -0.9999914169311523,
116
+ -0.9999884366989136,
117
+ -0.9975369572639465,
118
+ 6.754255821306288e-08
119
+ ],
120
+ "mean": [
121
+ -0.22290331318740714,
122
+ 0.04579037037503981,
123
+ -0.05804094906527823,
124
+ 0.12347219024917105
125
+ ],
126
+ "std": [
127
+ 0.8519267931761344,
128
+ 0.2949394732270141,
129
+ 0.30026078832556913,
130
+ 0.16333086331829588
131
+ ],
132
+ "q01": [
133
+ -0.9957959450418518,
134
+ -0.5678869493126506,
135
+ -0.5318135475761875,
136
+ 0.0014160465686576505
137
+ ],
138
+ "q99": [
139
+ 0.9900692296898317,
140
+ 0.6357358195673684,
141
+ 0.5009847868015971,
142
+ 0.4287044482210845
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05289580300450325,
148
+ 0.00937709677964449
149
+ ],
150
+ "min": [
151
+ -0.011669117026031017,
152
+ -0.06307756155729294
153
+ ],
154
+ "mean": [
155
+ 0.03203268930378034,
156
+ -0.032331639271968075
157
+ ],
158
+ "std": [
159
+ 0.011607049173153697,
160
+ 0.011306286745872529
161
+ ],
162
+ "q01": [
163
+ 0.006429109393656672,
164
+ -0.04056586943958638
165
+ ],
166
+ "q99": [
167
+ 0.040377860142451545,
168
+ -0.006593889698673258
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.006050336852760097,
188
+ 0.002728916466628117,
189
+ -0.0017236811393598838,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11032167711449951,
194
+ 0.09824140301615687,
195
+ 0.07825208047515658,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.154721262094005,
200
+ -0.18955302988697728,
201
+ -0.14454065867754942,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.24492955247744197,
206
+ 0.2496797822190429,
207
+ 0.11966742325336034,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9294448715752568
220
+ ],
221
+ "std": [
222
+ 0.36897903503358326
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.21627562513311385
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.00826780165926883,
244
+ 0.003161997359379232,
245
+ -0.08157453651841007
246
+ ],
247
+ "std": [
248
+ 0.4545430944464769,
249
+ 0.41040151434986594,
250
+ 0.4102644251536544
251
+ ],
252
+ "q01": [
253
+ -0.9653185255999107,
254
+ -0.9255635199318754,
255
+ -0.8208437313341718
256
+ ],
257
+ "q99": [
258
+ 0.7987781084592441,
259
+ 0.9260844232330088,
260
+ 0.8357421812622201
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.007367024641295137,
276
+ -0.02060756343728361,
277
+ -0.001884833075998489
278
+ ],
279
+ "std": [
280
+ 0.1116768956941076,
281
+ 0.13249042272165212,
282
+ 0.12338908364725322
283
+ ],
284
+ "q01": [
285
+ -0.26894658663246845,
286
+ -0.39865607740339326,
287
+ -0.34393735191302655
288
+ ],
289
+ "q99": [
290
+ 0.3191674780171587,
291
+ 0.31133630397431267,
292
+ 0.33968476423051097
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3098994894722959
304
+ ],
305
+ "std": [
306
+ 0.9507718287498876
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7545620899936146
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-3000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.33403685688972473,
9
+ 0.7223060727119446
10
+ ],
11
+ "min": [
12
+ -4.821934223175049,
13
+ -6.785110950469971,
14
+ 0.6994727253913879
15
+ ],
16
+ "mean": [
17
+ 2.5886094736594543,
18
+ -2.1559664577316586,
19
+ 0.7003580854548973
20
+ ],
21
+ "std": [
22
+ 1.6717149343420803,
23
+ 1.3743695399555862,
24
+ 0.000940069542359506
25
+ ],
26
+ "q01": [
27
+ -0.7517815060249613,
28
+ -5.497849509089705,
29
+ 0.6999776914424374
30
+ ],
31
+ "q99": [
32
+ 6.060945690807028,
33
+ -0.5331499932493624,
34
+ 0.7028864260472257
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.24488939171811902,
54
+ 0.5968549204077488
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6659970734507764,
60
+ 0.37443298392262264
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -0.9999999999999999,
66
+ 4.149924358336069e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 0.9999999999999999,
72
+ 0.9999999999999999
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8848381638526917,
78
+ 0.8569459915161133,
79
+ 1.053884744644165
80
+ ],
81
+ "min": [
82
+ -0.5828549861907959,
83
+ -0.8486236333847046,
84
+ -0.3446122705936432
85
+ ],
86
+ "mean": [
87
+ 0.25087486209693133,
88
+ -0.005738920462036393,
89
+ 0.43643948866361787
90
+ ],
91
+ "std": [
92
+ 0.15741544532518012,
93
+ 0.2547603106793105,
94
+ 0.2050434492495558
95
+ ],
96
+ "q01": [
97
+ -0.013990057288082561,
98
+ -0.4824599092623036,
99
+ 0.18825220353500022
100
+ ],
101
+ "q99": [
102
+ 0.5411989642841469,
103
+ 0.4754923657245658,
104
+ 0.6912323747882034
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999949932098389,
110
+ 0.9997354745864868,
111
+ 0.9900747537612915,
112
+ 0.9735639095306396
113
+ ],
114
+ "min": [
115
+ -0.9999914169311523,
116
+ -0.9999884366989136,
117
+ -0.9975369572639465,
118
+ 6.754255821306288e-08
119
+ ],
120
+ "mean": [
121
+ -0.22290331318740714,
122
+ 0.04579037037503981,
123
+ -0.05804094906527823,
124
+ 0.12347219024917105
125
+ ],
126
+ "std": [
127
+ 0.8519267931761344,
128
+ 0.2949394732270141,
129
+ 0.30026078832556913,
130
+ 0.16333086331829588
131
+ ],
132
+ "q01": [
133
+ -0.9957959450418518,
134
+ -0.5678869493126506,
135
+ -0.5318135475761875,
136
+ 0.0014160465686576505
137
+ ],
138
+ "q99": [
139
+ 0.9900692296898317,
140
+ 0.6357358195673684,
141
+ 0.5009847868015971,
142
+ 0.4287044482210845
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05289580300450325,
148
+ 0.00937709677964449
149
+ ],
150
+ "min": [
151
+ -0.011669117026031017,
152
+ -0.06307756155729294
153
+ ],
154
+ "mean": [
155
+ 0.03203268930378034,
156
+ -0.032331639271968075
157
+ ],
158
+ "std": [
159
+ 0.011607049173153697,
160
+ 0.011306286745872529
161
+ ],
162
+ "q01": [
163
+ 0.006429109393656672,
164
+ -0.04056586943958638
165
+ ],
166
+ "q99": [
167
+ 0.040377860142451545,
168
+ -0.006593889698673258
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.006050336852760097,
188
+ 0.002728916466628117,
189
+ -0.0017236811393598838,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11032167711449951,
194
+ 0.09824140301615687,
195
+ 0.07825208047515658,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.154721262094005,
200
+ -0.18955302988697728,
201
+ -0.14454065867754942,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.24492955247744197,
206
+ 0.2496797822190429,
207
+ 0.11966742325336034,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9294448715752568
220
+ ],
221
+ "std": [
222
+ 0.36897903503358326
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.21627562513311385
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.00826780165926883,
244
+ 0.003161997359379232,
245
+ -0.08157453651841007
246
+ ],
247
+ "std": [
248
+ 0.4545430944464769,
249
+ 0.41040151434986594,
250
+ 0.4102644251536544
251
+ ],
252
+ "q01": [
253
+ -0.9653185255999107,
254
+ -0.9255635199318754,
255
+ -0.8208437313341718
256
+ ],
257
+ "q99": [
258
+ 0.7987781084592441,
259
+ 0.9260844232330088,
260
+ 0.8357421812622201
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.007367024641295137,
276
+ -0.02060756343728361,
277
+ -0.001884833075998489
278
+ ],
279
+ "std": [
280
+ 0.1116768956941076,
281
+ 0.13249042272165212,
282
+ 0.12338908364725322
283
+ ],
284
+ "q01": [
285
+ -0.26894658663246845,
286
+ -0.39865607740339326,
287
+ -0.34393735191302655
288
+ ],
289
+ "q99": [
290
+ 0.3191674780171587,
291
+ 0.31133630397431267,
292
+ 0.33968476423051097
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3098994894722959
304
+ ],
305
+ "std": [
306
+ 0.9507718287498876
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7545620899936146
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/resolved_config.yaml ADDED
@@ -0,0 +1,38 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: fewshot_39tasks_5shot_from_rkd_timewarp_fullactionhead_seed1
2
+ policy_type: groot_RKDv3
3
+ base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1
5
+ dataset:
6
+ dataset_soup: fewshot_heldout_atomic39_human_5shot
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 64
10
+ seed: 42
11
+ filter_key_seed: 1
12
+ manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed1.json
13
+ save_total_limit: 4
14
+ model:
15
+ rkd_enabled: true
16
+ rkd_fm_loss_weight: 1.0
17
+ rkd_loss_weight: 0.5
18
+ rkd_loss_weight_schedule: null
19
+ phases:
20
+ - name: train_5shot_seed1_fullactionhead_5k
21
+ max_steps: 5000
22
+ save_steps: 0
23
+ save_step_list:
24
+ - 1000
25
+ - 3000
26
+ - 5000
27
+ save_final_model: false
28
+ trainable:
29
+ tune_llm: false
30
+ tune_visual: false
31
+ tune_projector: true
32
+ tune_diffusion_model: true
33
+ losses:
34
+ rkd_enabled: true
35
+ rkd_fm_loss_weight: 1.0
36
+ rkd_loss_weight: 0.5
37
+ rkd_loss_weight_schedule: null
38
+ resolved_sweep: {}
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/seed1.yaml ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: fewshot_39tasks_5shot_from_rkd_timewarp_fullactionhead_seed1
2
+ policy_type: groot_RKDv3
3
+ base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1
5
+ dataset:
6
+ dataset_soup: fewshot_heldout_atomic39_human_5shot
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 64
10
+ seed: 42
11
+ filter_key_seed: 1
12
+ manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed1.json
13
+ save_total_limit: 4
14
+ model:
15
+ rkd_enabled: true
16
+ rkd_fm_loss_weight: 1.0
17
+ rkd_loss_weight: 0.5
18
+ rkd_loss_weight_schedule: null
19
+ phases:
20
+ - name: train_5shot_seed1_fullactionhead_5k
21
+ max_steps: 5000
22
+ save_steps: 0
23
+ save_step_list:
24
+ - 1000
25
+ - 3000
26
+ - 5000
27
+ save_final_model: false
28
+ trainable:
29
+ tune_llm: false
30
+ tune_visual: false
31
+ tune_projector: true
32
+ tune_diffusion_model: true
33
+ losses:
34
+ rkd_enabled: true
35
+ rkd_fm_loss_weight: 1.0
36
+ rkd_loss_weight: 0.5
37
+ rkd_loss_weight_schedule: null
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/config.json ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_encoder_cls_ffn_hidden_dim": 2048,
63
+ "rkd_action_encoder_cls_num_heads": 8,
64
+ "rkd_action_encoder_cls_num_layers": 1,
65
+ "rkd_action_encoder_projector_dim": 512,
66
+ "rkd_action_encoder_projector_enabled": false,
67
+ "rkd_action_encoder_projector_pooling": "flatten",
68
+ "rkd_action_temp": 0.1,
69
+ "rkd_angle_loss_weight": 2.0,
70
+ "rkd_distance_loss_weight": 1.0,
71
+ "rkd_dtw_adaptive_floor": 0.05,
72
+ "rkd_dtw_adaptive_mix_lambda": 0.5,
73
+ "rkd_dtw_base_weight": 1.0,
74
+ "rkd_dtw_gripper_weight": 0.5,
75
+ "rkd_dtw_mix_mode": "distance_mix",
76
+ "rkd_dtw_normalize_components": true,
77
+ "rkd_dtw_pos_weight": 1.0,
78
+ "rkd_dtw_rot_weight": 1.0,
79
+ "rkd_dtw_timing_log_interval": 1000,
80
+ "rkd_enabled": true,
81
+ "rkd_exclude_diagonal": true,
82
+ "rkd_fm_loss_weight": 1.0,
83
+ "rkd_loss_type": "distance_angle",
84
+ "rkd_loss_weight": 0.5,
85
+ "rkd_loss_weight_end": 0.0,
86
+ "rkd_loss_weight_schedule": null,
87
+ "rkd_loss_weight_start": 0.1,
88
+ "rkd_raw_action_projector_dim": 512,
89
+ "rkd_raw_action_projector_enabled": false,
90
+ "rkd_raw_action_projector_pooling": "flatten",
91
+ "rkd_relation_mode": "token_pair_mean",
92
+ "rkd_student_reconstruction_head_dim": 512,
93
+ "rkd_student_reconstruction_head_enabled": true,
94
+ "rkd_student_reconstruction_head_hidden_dim": 512,
95
+ "rkd_student_reconstruction_head_pooling": "attention",
96
+ "rkd_teacher_source": "timewarp_action",
97
+ "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
98
+ "rkd_vlm_temp": 0.1,
99
+ "torch_dtype": "bfloat16",
100
+ "transformers_version": "4.51.3"
101
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.33403685688972473,
9
+ 0.7223060727119446
10
+ ],
11
+ "min": [
12
+ -4.821934223175049,
13
+ -6.785110950469971,
14
+ 0.6994727253913879
15
+ ],
16
+ "mean": [
17
+ 2.5719731235192254,
18
+ -2.1489371002267235,
19
+ 0.7003612741268546
20
+ ],
21
+ "std": [
22
+ 1.669945884010784,
23
+ 1.3739411381361726,
24
+ 0.0009413282367754987
25
+ ],
26
+ "q01": [
27
+ -0.7944286339399775,
28
+ -5.486611518471242,
29
+ 0.699977555971739
30
+ ],
31
+ "q99": [
32
+ 6.019886905263648,
33
+ -0.5229184210026855,
34
+ 0.702930326282967
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.248543484335273,
54
+ 0.5959317405459531
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6655550240074092,
60
+ 0.374283995143559
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -1.0000000000000002,
66
+ 4.503254968910557e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 1.0000000000000002,
72
+ 1.0000000000000002
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8848381638526917,
78
+ 0.8569459915161133,
79
+ 1.053884744644165
80
+ ],
81
+ "min": [
82
+ -0.5828549861907959,
83
+ -0.8486236333847046,
84
+ -0.3446122705936432
85
+ ],
86
+ "mean": [
87
+ 0.2465490874130862,
88
+ -0.004711658908481467,
89
+ 0.43463835464312073
90
+ ],
91
+ "std": [
92
+ 0.15798784670671595,
93
+ 0.25624102400497073,
94
+ 0.20836028727974915
95
+ ],
96
+ "q01": [
97
+ -0.020752051134640608,
98
+ -0.4837933887347875,
99
+ 0.1802942952289915
100
+ ],
101
+ "q99": [
102
+ 0.5386784714844007,
103
+ 0.4801424212024721,
104
+ 0.695095732975766
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999949932098389,
110
+ 0.9997354745864868,
111
+ 0.9900747537612915,
112
+ 0.9735639095306396
113
+ ],
114
+ "min": [
115
+ -0.9999914169311523,
116
+ -0.9999884366989136,
117
+ -0.9975369572639465,
118
+ 6.754255821306288e-08
119
+ ],
120
+ "mean": [
121
+ -0.22531859714056926,
122
+ 0.046973702821026614,
123
+ -0.0568176967878771,
124
+ 0.12364307042955697
125
+ ],
126
+ "std": [
127
+ 0.8502144333612052,
128
+ 0.29591250318147405,
129
+ 0.30218053156837016,
130
+ 0.1636031353412931
131
+ ],
132
+ "q01": [
133
+ -0.9959057387172566,
134
+ -0.5665384745573048,
135
+ -0.5329334620834977,
136
+ 0.0014372279999397553
137
+ ],
138
+ "q99": [
139
+ 0.9904107951565839,
140
+ 0.6359623612021396,
141
+ 0.5026076022439377,
142
+ 0.42955559234879914
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05289580300450325,
148
+ 0.00937709677964449
149
+ ],
150
+ "min": [
151
+ -0.011669117026031017,
152
+ -0.06307756155729294
153
+ ],
154
+ "mean": [
155
+ 0.03188757060429838,
156
+ -0.0321793333086053
157
+ ],
158
+ "std": [
159
+ 0.01171052313737148,
160
+ 0.011427336801203189
161
+ ],
162
+ "q01": [
163
+ 0.006462901319459925,
164
+ -0.040557140599419166
165
+ ],
166
+ "q99": [
167
+ 0.04037619335388386,
168
+ -0.006523947835993347
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.006639007355857881,
188
+ 0.0027330421795561874,
189
+ -0.0017774441970214892,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11841528114040736,
194
+ 0.1041359083946544,
195
+ 0.08346265521208351,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.17562359675270237,
200
+ -0.21706402321529558,
201
+ -0.16273392585723104,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.2572918701644887,
206
+ 0.27022325575787115,
207
+ 0.1369626751049278,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9205539608927732
220
+ ],
221
+ "std": [
222
+ 0.39063657242534666
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.17711863866562935
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.00924803964621179,
244
+ 0.004053841761390575,
245
+ -0.08385115578460103
246
+ ],
247
+ "std": [
248
+ 0.45413072518223824,
249
+ 0.40158612714634856,
250
+ 0.41068634292849304
251
+ ],
252
+ "q01": [
253
+ -0.9669817145875343,
254
+ -0.9241307799783718,
255
+ -0.8264079503068825
256
+ ],
257
+ "q99": [
258
+ 0.8021511975722166,
259
+ 0.9241346582748478,
260
+ 0.8348116781502298
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.007644157504353809,
276
+ -0.019962543093886335,
277
+ -0.002047110452602617
278
+ ],
279
+ "std": [
280
+ 0.11076012154791087,
281
+ 0.13176735140285795,
282
+ 0.12286528665895614
283
+ ],
284
+ "q01": [
285
+ -0.2659105439401192,
286
+ -0.3961694033813533,
287
+ -0.3434870832468936
288
+ ],
289
+ "q99": [
290
+ 0.3145786959266369,
291
+ 0.31309430361227547,
292
+ 0.34107754401942986
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.29870612267533575
304
+ ],
305
+ "std": [
306
+ 0.9543487555375691
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7610674802484976
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-1000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/config.json ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_encoder_cls_ffn_hidden_dim": 2048,
63
+ "rkd_action_encoder_cls_num_heads": 8,
64
+ "rkd_action_encoder_cls_num_layers": 1,
65
+ "rkd_action_encoder_projector_dim": 512,
66
+ "rkd_action_encoder_projector_enabled": false,
67
+ "rkd_action_encoder_projector_pooling": "flatten",
68
+ "rkd_action_temp": 0.1,
69
+ "rkd_angle_loss_weight": 2.0,
70
+ "rkd_distance_loss_weight": 1.0,
71
+ "rkd_dtw_adaptive_floor": 0.05,
72
+ "rkd_dtw_adaptive_mix_lambda": 0.5,
73
+ "rkd_dtw_base_weight": 1.0,
74
+ "rkd_dtw_gripper_weight": 0.5,
75
+ "rkd_dtw_mix_mode": "distance_mix",
76
+ "rkd_dtw_normalize_components": true,
77
+ "rkd_dtw_pos_weight": 1.0,
78
+ "rkd_dtw_rot_weight": 1.0,
79
+ "rkd_dtw_timing_log_interval": 1000,
80
+ "rkd_enabled": true,
81
+ "rkd_exclude_diagonal": true,
82
+ "rkd_fm_loss_weight": 1.0,
83
+ "rkd_loss_type": "distance_angle",
84
+ "rkd_loss_weight": 0.5,
85
+ "rkd_loss_weight_end": 0.0,
86
+ "rkd_loss_weight_schedule": null,
87
+ "rkd_loss_weight_start": 0.1,
88
+ "rkd_raw_action_projector_dim": 512,
89
+ "rkd_raw_action_projector_enabled": false,
90
+ "rkd_raw_action_projector_pooling": "flatten",
91
+ "rkd_relation_mode": "token_pair_mean",
92
+ "rkd_student_reconstruction_head_dim": 512,
93
+ "rkd_student_reconstruction_head_enabled": true,
94
+ "rkd_student_reconstruction_head_hidden_dim": 512,
95
+ "rkd_student_reconstruction_head_pooling": "attention",
96
+ "rkd_teacher_source": "timewarp_action",
97
+ "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
98
+ "rkd_vlm_temp": 0.1,
99
+ "torch_dtype": "bfloat16",
100
+ "transformers_version": "4.51.3"
101
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.33403685688972473,
9
+ 0.7223060727119446
10
+ ],
11
+ "min": [
12
+ -4.821934223175049,
13
+ -6.785110950469971,
14
+ 0.6994727253913879
15
+ ],
16
+ "mean": [
17
+ 2.5719731235192254,
18
+ -2.1489371002267235,
19
+ 0.7003612741268546
20
+ ],
21
+ "std": [
22
+ 1.669945884010784,
23
+ 1.3739411381361726,
24
+ 0.0009413282367754987
25
+ ],
26
+ "q01": [
27
+ -0.7944286339399775,
28
+ -5.486611518471242,
29
+ 0.699977555971739
30
+ ],
31
+ "q99": [
32
+ 6.019886905263648,
33
+ -0.5229184210026855,
34
+ 0.702930326282967
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.248543484335273,
54
+ 0.5959317405459531
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6655550240074092,
60
+ 0.374283995143559
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -1.0000000000000002,
66
+ 4.503254968910557e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 1.0000000000000002,
72
+ 1.0000000000000002
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8848381638526917,
78
+ 0.8569459915161133,
79
+ 1.053884744644165
80
+ ],
81
+ "min": [
82
+ -0.5828549861907959,
83
+ -0.8486236333847046,
84
+ -0.3446122705936432
85
+ ],
86
+ "mean": [
87
+ 0.2465490874130862,
88
+ -0.004711658908481467,
89
+ 0.43463835464312073
90
+ ],
91
+ "std": [
92
+ 0.15798784670671595,
93
+ 0.25624102400497073,
94
+ 0.20836028727974915
95
+ ],
96
+ "q01": [
97
+ -0.020752051134640608,
98
+ -0.4837933887347875,
99
+ 0.1802942952289915
100
+ ],
101
+ "q99": [
102
+ 0.5386784714844007,
103
+ 0.4801424212024721,
104
+ 0.695095732975766
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999949932098389,
110
+ 0.9997354745864868,
111
+ 0.9900747537612915,
112
+ 0.9735639095306396
113
+ ],
114
+ "min": [
115
+ -0.9999914169311523,
116
+ -0.9999884366989136,
117
+ -0.9975369572639465,
118
+ 6.754255821306288e-08
119
+ ],
120
+ "mean": [
121
+ -0.22531859714056926,
122
+ 0.046973702821026614,
123
+ -0.0568176967878771,
124
+ 0.12364307042955697
125
+ ],
126
+ "std": [
127
+ 0.8502144333612052,
128
+ 0.29591250318147405,
129
+ 0.30218053156837016,
130
+ 0.1636031353412931
131
+ ],
132
+ "q01": [
133
+ -0.9959057387172566,
134
+ -0.5665384745573048,
135
+ -0.5329334620834977,
136
+ 0.0014372279999397553
137
+ ],
138
+ "q99": [
139
+ 0.9904107951565839,
140
+ 0.6359623612021396,
141
+ 0.5026076022439377,
142
+ 0.42955559234879914
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05289580300450325,
148
+ 0.00937709677964449
149
+ ],
150
+ "min": [
151
+ -0.011669117026031017,
152
+ -0.06307756155729294
153
+ ],
154
+ "mean": [
155
+ 0.03188757060429838,
156
+ -0.0321793333086053
157
+ ],
158
+ "std": [
159
+ 0.01171052313737148,
160
+ 0.011427336801203189
161
+ ],
162
+ "q01": [
163
+ 0.006462901319459925,
164
+ -0.040557140599419166
165
+ ],
166
+ "q99": [
167
+ 0.04037619335388386,
168
+ -0.006523947835993347
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.006639007355857881,
188
+ 0.0027330421795561874,
189
+ -0.0017774441970214892,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11841528114040736,
194
+ 0.1041359083946544,
195
+ 0.08346265521208351,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.17562359675270237,
200
+ -0.21706402321529558,
201
+ -0.16273392585723104,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.2572918701644887,
206
+ 0.27022325575787115,
207
+ 0.1369626751049278,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9205539608927732
220
+ ],
221
+ "std": [
222
+ 0.39063657242534666
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.17711863866562935
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.00924803964621179,
244
+ 0.004053841761390575,
245
+ -0.08385115578460103
246
+ ],
247
+ "std": [
248
+ 0.45413072518223824,
249
+ 0.40158612714634856,
250
+ 0.41068634292849304
251
+ ],
252
+ "q01": [
253
+ -0.9669817145875343,
254
+ -0.9241307799783718,
255
+ -0.8264079503068825
256
+ ],
257
+ "q99": [
258
+ 0.8021511975722166,
259
+ 0.9241346582748478,
260
+ 0.8348116781502298
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.007644157504353809,
276
+ -0.019962543093886335,
277
+ -0.002047110452602617
278
+ ],
279
+ "std": [
280
+ 0.11076012154791087,
281
+ 0.13176735140285795,
282
+ 0.12286528665895614
283
+ ],
284
+ "q01": [
285
+ -0.2659105439401192,
286
+ -0.3961694033813533,
287
+ -0.3434870832468936
288
+ ],
289
+ "q99": [
290
+ 0.3145786959266369,
291
+ 0.31309430361227547,
292
+ 0.34107754401942986
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.29870612267533575
304
+ ],
305
+ "std": [
306
+ 0.9543487555375691
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7610674802484976
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-3000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.33403685688972473,
9
+ 0.7223060727119446
10
+ ],
11
+ "min": [
12
+ -4.821934223175049,
13
+ -6.785110950469971,
14
+ 0.6994727253913879
15
+ ],
16
+ "mean": [
17
+ 2.5719731235192254,
18
+ -2.1489371002267235,
19
+ 0.7003612741268546
20
+ ],
21
+ "std": [
22
+ 1.669945884010784,
23
+ 1.3739411381361726,
24
+ 0.0009413282367754987
25
+ ],
26
+ "q01": [
27
+ -0.7944286339399775,
28
+ -5.486611518471242,
29
+ 0.699977555971739
30
+ ],
31
+ "q99": [
32
+ 6.019886905263648,
33
+ -0.5229184210026855,
34
+ 0.702930326282967
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.248543484335273,
54
+ 0.5959317405459531
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6655550240074092,
60
+ 0.374283995143559
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -1.0000000000000002,
66
+ 4.503254968910557e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 1.0000000000000002,
72
+ 1.0000000000000002
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8848381638526917,
78
+ 0.8569459915161133,
79
+ 1.053884744644165
80
+ ],
81
+ "min": [
82
+ -0.5828549861907959,
83
+ -0.8486236333847046,
84
+ -0.3446122705936432
85
+ ],
86
+ "mean": [
87
+ 0.2465490874130862,
88
+ -0.004711658908481467,
89
+ 0.43463835464312073
90
+ ],
91
+ "std": [
92
+ 0.15798784670671595,
93
+ 0.25624102400497073,
94
+ 0.20836028727974915
95
+ ],
96
+ "q01": [
97
+ -0.020752051134640608,
98
+ -0.4837933887347875,
99
+ 0.1802942952289915
100
+ ],
101
+ "q99": [
102
+ 0.5386784714844007,
103
+ 0.4801424212024721,
104
+ 0.695095732975766
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999949932098389,
110
+ 0.9997354745864868,
111
+ 0.9900747537612915,
112
+ 0.9735639095306396
113
+ ],
114
+ "min": [
115
+ -0.9999914169311523,
116
+ -0.9999884366989136,
117
+ -0.9975369572639465,
118
+ 6.754255821306288e-08
119
+ ],
120
+ "mean": [
121
+ -0.22531859714056926,
122
+ 0.046973702821026614,
123
+ -0.0568176967878771,
124
+ 0.12364307042955697
125
+ ],
126
+ "std": [
127
+ 0.8502144333612052,
128
+ 0.29591250318147405,
129
+ 0.30218053156837016,
130
+ 0.1636031353412931
131
+ ],
132
+ "q01": [
133
+ -0.9959057387172566,
134
+ -0.5665384745573048,
135
+ -0.5329334620834977,
136
+ 0.0014372279999397553
137
+ ],
138
+ "q99": [
139
+ 0.9904107951565839,
140
+ 0.6359623612021396,
141
+ 0.5026076022439377,
142
+ 0.42955559234879914
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05289580300450325,
148
+ 0.00937709677964449
149
+ ],
150
+ "min": [
151
+ -0.011669117026031017,
152
+ -0.06307756155729294
153
+ ],
154
+ "mean": [
155
+ 0.03188757060429838,
156
+ -0.0321793333086053
157
+ ],
158
+ "std": [
159
+ 0.01171052313737148,
160
+ 0.011427336801203189
161
+ ],
162
+ "q01": [
163
+ 0.006462901319459925,
164
+ -0.040557140599419166
165
+ ],
166
+ "q99": [
167
+ 0.04037619335388386,
168
+ -0.006523947835993347
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.006639007355857881,
188
+ 0.0027330421795561874,
189
+ -0.0017774441970214892,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11841528114040736,
194
+ 0.1041359083946544,
195
+ 0.08346265521208351,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.17562359675270237,
200
+ -0.21706402321529558,
201
+ -0.16273392585723104,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.2572918701644887,
206
+ 0.27022325575787115,
207
+ 0.1369626751049278,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9205539608927732
220
+ ],
221
+ "std": [
222
+ 0.39063657242534666
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.17711863866562935
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.00924803964621179,
244
+ 0.004053841761390575,
245
+ -0.08385115578460103
246
+ ],
247
+ "std": [
248
+ 0.45413072518223824,
249
+ 0.40158612714634856,
250
+ 0.41068634292849304
251
+ ],
252
+ "q01": [
253
+ -0.9669817145875343,
254
+ -0.9241307799783718,
255
+ -0.8264079503068825
256
+ ],
257
+ "q99": [
258
+ 0.8021511975722166,
259
+ 0.9241346582748478,
260
+ 0.8348116781502298
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.007644157504353809,
276
+ -0.019962543093886335,
277
+ -0.002047110452602617
278
+ ],
279
+ "std": [
280
+ 0.11076012154791087,
281
+ 0.13176735140285795,
282
+ 0.12286528665895614
283
+ ],
284
+ "q01": [
285
+ -0.2659105439401192,
286
+ -0.3961694033813533,
287
+ -0.3434870832468936
288
+ ],
289
+ "q99": [
290
+ 0.3145786959266369,
291
+ 0.31309430361227547,
292
+ 0.34107754401942986
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.29870612267533575
304
+ ],
305
+ "std": [
306
+ 0.9543487555375691
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7610674802484976
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/resolved_config.yaml ADDED
@@ -0,0 +1,38 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: fewshot_39tasks_5shot_from_rkd_timewarp_fullactionhead_seed2
2
+ policy_type: groot_RKDv3
3
+ base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2
5
+ dataset:
6
+ dataset_soup: fewshot_heldout_atomic39_human_5shot
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 64
10
+ seed: 42
11
+ filter_key_seed: 2
12
+ manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed2.json
13
+ save_total_limit: 4
14
+ model:
15
+ rkd_enabled: true
16
+ rkd_fm_loss_weight: 1.0
17
+ rkd_loss_weight: 0.5
18
+ rkd_loss_weight_schedule: null
19
+ phases:
20
+ - name: train_5shot_seed2_fullactionhead_5k
21
+ max_steps: 5000
22
+ save_steps: 0
23
+ save_step_list:
24
+ - 1000
25
+ - 3000
26
+ - 5000
27
+ save_final_model: false
28
+ trainable:
29
+ tune_llm: false
30
+ tune_visual: false
31
+ tune_projector: true
32
+ tune_diffusion_model: true
33
+ losses:
34
+ rkd_enabled: true
35
+ rkd_fm_loss_weight: 1.0
36
+ rkd_loss_weight: 0.5
37
+ rkd_loss_weight_schedule: null
38
+ resolved_sweep: {}
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/seed2.yaml ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ experiment_name: fewshot_39tasks_5shot_from_rkd_timewarp_fullactionhead_seed2
2
+ policy_type: groot_RKDv3
3
+ base_model_path: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_from_Atomic26Baseline60k/default/phase3
4
+ output_root: ${HOME}/groot_robocasa/robocasa_v2/FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2
5
+ dataset:
6
+ dataset_soup: fewshot_heldout_atomic39_human_5shot
7
+ training:
8
+ num_gpus: 1
9
+ batch_size: 64
10
+ seed: 42
11
+ filter_key_seed: 2
12
+ manifest_path: ${HOME}/clvla/benchmarks/robocasa_v2/my_scripts/Fewshot/manifests/fewshot_heldout_atomic39/shot5_seed2.json
13
+ save_total_limit: 4
14
+ model:
15
+ rkd_enabled: true
16
+ rkd_fm_loss_weight: 1.0
17
+ rkd_loss_weight: 0.5
18
+ rkd_loss_weight_schedule: null
19
+ phases:
20
+ - name: train_5shot_seed2_fullactionhead_5k
21
+ max_steps: 5000
22
+ save_steps: 0
23
+ save_step_list:
24
+ - 1000
25
+ - 3000
26
+ - 5000
27
+ save_final_model: false
28
+ trainable:
29
+ tune_llm: false
30
+ tune_visual: false
31
+ tune_projector: true
32
+ tune_diffusion_model: true
33
+ losses:
34
+ rkd_enabled: true
35
+ rkd_fm_loss_weight: 1.0
36
+ rkd_loss_weight: 0.5
37
+ rkd_loss_weight_schedule: null
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/config.json ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_encoder_cls_ffn_hidden_dim": 2048,
63
+ "rkd_action_encoder_cls_num_heads": 8,
64
+ "rkd_action_encoder_cls_num_layers": 1,
65
+ "rkd_action_encoder_projector_dim": 512,
66
+ "rkd_action_encoder_projector_enabled": false,
67
+ "rkd_action_encoder_projector_pooling": "flatten",
68
+ "rkd_action_temp": 0.1,
69
+ "rkd_angle_loss_weight": 2.0,
70
+ "rkd_distance_loss_weight": 1.0,
71
+ "rkd_dtw_adaptive_floor": 0.05,
72
+ "rkd_dtw_adaptive_mix_lambda": 0.5,
73
+ "rkd_dtw_base_weight": 1.0,
74
+ "rkd_dtw_gripper_weight": 0.5,
75
+ "rkd_dtw_mix_mode": "distance_mix",
76
+ "rkd_dtw_normalize_components": true,
77
+ "rkd_dtw_pos_weight": 1.0,
78
+ "rkd_dtw_rot_weight": 1.0,
79
+ "rkd_dtw_timing_log_interval": 1000,
80
+ "rkd_enabled": true,
81
+ "rkd_exclude_diagonal": true,
82
+ "rkd_fm_loss_weight": 1.0,
83
+ "rkd_loss_type": "distance_angle",
84
+ "rkd_loss_weight": 0.5,
85
+ "rkd_loss_weight_end": 0.0,
86
+ "rkd_loss_weight_schedule": null,
87
+ "rkd_loss_weight_start": 0.1,
88
+ "rkd_raw_action_projector_dim": 512,
89
+ "rkd_raw_action_projector_enabled": false,
90
+ "rkd_raw_action_projector_pooling": "flatten",
91
+ "rkd_relation_mode": "token_pair_mean",
92
+ "rkd_student_reconstruction_head_dim": 512,
93
+ "rkd_student_reconstruction_head_enabled": true,
94
+ "rkd_student_reconstruction_head_hidden_dim": 512,
95
+ "rkd_student_reconstruction_head_pooling": "attention",
96
+ "rkd_teacher_source": "timewarp_action",
97
+ "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
98
+ "rkd_vlm_temp": 0.1,
99
+ "torch_dtype": "bfloat16",
100
+ "transformers_version": "4.51.3"
101
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.33403685688972473,
9
+ 0.7223060727119446
10
+ ],
11
+ "min": [
12
+ -4.821934223175049,
13
+ -6.785110950469971,
14
+ 0.6994727253913879
15
+ ],
16
+ "mean": [
17
+ 2.577927505176127,
18
+ -2.1525618086425604,
19
+ 0.7003541659101092
20
+ ],
21
+ "std": [
22
+ 1.6720365948940972,
23
+ 1.3768414767482617,
24
+ 0.0009102921174316669
25
+ ],
26
+ "q01": [
27
+ -0.7605057497562686,
28
+ -5.518671947805291,
29
+ 0.6999764703640262
30
+ ],
31
+ "q99": [
32
+ 6.038879433964306,
33
+ -0.5168299981147954,
34
+ 0.7029182461269848
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.24782371371751144,
54
+ 0.5971093380149053
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6642559499983499,
60
+ 0.3751891785184018
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -1.0,
66
+ 4.194193576416142e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 1.0,
72
+ 1.0
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8848381638526917,
78
+ 0.8569459915161133,
79
+ 1.053884744644165
80
+ ],
81
+ "min": [
82
+ -0.5828549861907959,
83
+ -0.8486236333847046,
84
+ -0.3446122705936432
85
+ ],
86
+ "mean": [
87
+ 0.25006971579930243,
88
+ -0.005716176032514759,
89
+ 0.4426027721228362
90
+ ],
91
+ "std": [
92
+ 0.15838384751435522,
93
+ 0.25141504139800513,
94
+ 0.20646816234620438
95
+ ],
96
+ "q01": [
97
+ -0.017003721077380903,
98
+ -0.47738320165276543,
99
+ 0.1917563409753314
100
+ ],
101
+ "q99": [
102
+ 0.5401029199137674,
103
+ 0.4706646609268599,
104
+ 0.6965586537529516
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999949932098389,
110
+ 0.9997354745864868,
111
+ 0.9900747537612915,
112
+ 0.9735639095306396
113
+ ],
114
+ "min": [
115
+ -0.9999914169311523,
116
+ -0.9999884366989136,
117
+ -0.9975369572639465,
118
+ 6.754255821306288e-08
119
+ ],
120
+ "mean": [
121
+ -0.22531628862537245,
122
+ 0.045669797538445966,
123
+ -0.06098055586474368,
124
+ 0.12292005493282424
125
+ ],
126
+ "std": [
127
+ 0.8506341437448487,
128
+ 0.2880763358322842,
129
+ 0.30816573781739026,
130
+ 0.16366165154486362
131
+ ],
132
+ "q01": [
133
+ -0.9957510593732032,
134
+ -0.5584741016891311,
135
+ -0.5371647381159943,
136
+ 0.0013784815838340492
137
+ ],
138
+ "q99": [
139
+ 0.9899375209218226,
140
+ 0.6327878884078038,
141
+ 0.5071539024531994,
142
+ 0.43555975763914345
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05289580300450325,
148
+ 0.00937709677964449
149
+ ],
150
+ "min": [
151
+ -0.011669117026031017,
152
+ -0.06307756155729294
153
+ ],
154
+ "mean": [
155
+ 0.03196506153371015,
156
+ -0.03229791582989581
157
+ ],
158
+ "std": [
159
+ 0.01171640503534049,
160
+ 0.01140016755425199
161
+ ],
162
+ "q01": [
163
+ 0.006290779996034005,
164
+ -0.04059717954551848
165
+ ],
166
+ "q99": [
167
+ 0.04038165838477291,
168
+ -0.006363076424932449
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.006112990930475949,
188
+ 0.002504955933349057,
189
+ -0.001639409958610569,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11478077265783658,
194
+ 0.1002662368589144,
195
+ 0.0807258815139021,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.16969970921613756,
200
+ -0.19919537649977875,
201
+ -0.15012139866371593,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.2466625453958827,
206
+ 0.25040842772679267,
207
+ 0.12777701094549537,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9253067479599785
220
+ ],
221
+ "std": [
222
+ 0.3792397014797737
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.1974093068624371
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.006754222051360652,
244
+ 0.0027854853184013867,
245
+ -0.08087572130330477
246
+ ],
247
+ "std": [
248
+ 0.45779905420789546,
249
+ 0.40578975219522756,
250
+ 0.4062694098132682
251
+ ],
252
+ "q01": [
253
+ -0.968295960290036,
254
+ -0.9240402337368848,
255
+ -0.8249359802202183
256
+ ],
257
+ "q99": [
258
+ 0.8046634915851651,
259
+ 0.9212243025654695,
260
+ 0.838433711437593
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.008633749251857347,
276
+ -0.020406457524911718,
277
+ -0.0027639446362883006
278
+ ],
279
+ "std": [
280
+ 0.11393972872407104,
281
+ 0.13300186268009506,
282
+ 0.12192288302920105
283
+ ],
284
+ "q01": [
285
+ -0.2668579353776187,
286
+ -0.40299251206771697,
287
+ -0.34712957892172586
288
+ ],
289
+ "q99": [
290
+ 0.32748281264383927,
291
+ 0.3179028969292779,
292
+ 0.33459405510046736
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3091635769453286
304
+ ],
305
+ "std": [
306
+ 0.9510122432876397
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7643545615216344
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-1000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/config.json ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_dim": 32,
3
+ "action_head_cfg": {
4
+ "action_dim": 32,
5
+ "action_horizon": 16,
6
+ "add_pos_embed": true,
7
+ "backbone_embedding_dim": 2048,
8
+ "diffusion_model_cfg": {
9
+ "attention_head_dim": 48,
10
+ "cross_attention_dim": 2048,
11
+ "dropout": 0.2,
12
+ "final_dropout": true,
13
+ "interleave_self_attention": true,
14
+ "norm_type": "ada_norm",
15
+ "num_attention_heads": 32,
16
+ "num_layers": 16,
17
+ "output_dim": 1024,
18
+ "positional_embeddings": null
19
+ },
20
+ "hidden_size": 1024,
21
+ "input_embedding_dim": 1536,
22
+ "max_action_dim": 32,
23
+ "max_state_dim": 64,
24
+ "model_dtype": "float32",
25
+ "noise_beta_alpha": 1.5,
26
+ "noise_beta_beta": 1.0,
27
+ "noise_s": 0.999,
28
+ "num_inference_timesteps": 4,
29
+ "num_target_vision_tokens": 32,
30
+ "num_timestep_buckets": 1000,
31
+ "tune_diffusion_model": true,
32
+ "tune_projector": true,
33
+ "use_vlln": true,
34
+ "vl_self_attention_cfg": {
35
+ "attention_head_dim": 64,
36
+ "dropout": 0.2,
37
+ "final_dropout": true,
38
+ "num_attention_heads": 32,
39
+ "num_layers": 4,
40
+ "positional_embeddings": null
41
+ }
42
+ },
43
+ "action_horizon": 16,
44
+ "architectures": [
45
+ "GR00T_N1_5_RKD"
46
+ ],
47
+ "attn_implementation": null,
48
+ "backbone_cfg": {
49
+ "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
50
+ "load_bf16": false,
51
+ "project_to_dim": null,
52
+ "reproject_vision": false,
53
+ "select_layer": 12,
54
+ "tune_llm": false,
55
+ "tune_visual": true,
56
+ "use_flash_attention": true
57
+ },
58
+ "compute_dtype": "bfloat16",
59
+ "hidden_size": 2048,
60
+ "model_dtype": "float32",
61
+ "model_type": "gr00t_n1_5",
62
+ "rkd_action_encoder_cls_ffn_hidden_dim": 2048,
63
+ "rkd_action_encoder_cls_num_heads": 8,
64
+ "rkd_action_encoder_cls_num_layers": 1,
65
+ "rkd_action_encoder_projector_dim": 512,
66
+ "rkd_action_encoder_projector_enabled": false,
67
+ "rkd_action_encoder_projector_pooling": "flatten",
68
+ "rkd_action_temp": 0.1,
69
+ "rkd_angle_loss_weight": 2.0,
70
+ "rkd_distance_loss_weight": 1.0,
71
+ "rkd_dtw_adaptive_floor": 0.05,
72
+ "rkd_dtw_adaptive_mix_lambda": 0.5,
73
+ "rkd_dtw_base_weight": 1.0,
74
+ "rkd_dtw_gripper_weight": 0.5,
75
+ "rkd_dtw_mix_mode": "distance_mix",
76
+ "rkd_dtw_normalize_components": true,
77
+ "rkd_dtw_pos_weight": 1.0,
78
+ "rkd_dtw_rot_weight": 1.0,
79
+ "rkd_dtw_timing_log_interval": 1000,
80
+ "rkd_enabled": true,
81
+ "rkd_exclude_diagonal": true,
82
+ "rkd_fm_loss_weight": 1.0,
83
+ "rkd_loss_type": "distance_angle",
84
+ "rkd_loss_weight": 0.5,
85
+ "rkd_loss_weight_end": 0.0,
86
+ "rkd_loss_weight_schedule": null,
87
+ "rkd_loss_weight_start": 0.1,
88
+ "rkd_raw_action_projector_dim": 512,
89
+ "rkd_raw_action_projector_enabled": false,
90
+ "rkd_raw_action_projector_pooling": "flatten",
91
+ "rkd_relation_mode": "token_pair_mean",
92
+ "rkd_student_reconstruction_head_dim": 512,
93
+ "rkd_student_reconstruction_head_enabled": true,
94
+ "rkd_student_reconstruction_head_hidden_dim": 512,
95
+ "rkd_student_reconstruction_head_pooling": "attention",
96
+ "rkd_teacher_source": "timewarp_action",
97
+ "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
98
+ "rkd_vlm_temp": 0.1,
99
+ "torch_dtype": "bfloat16",
100
+ "transformers_version": "4.51.3"
101
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.33403685688972473,
9
+ 0.7223060727119446
10
+ ],
11
+ "min": [
12
+ -4.821934223175049,
13
+ -6.785110950469971,
14
+ 0.6994727253913879
15
+ ],
16
+ "mean": [
17
+ 2.577927505176127,
18
+ -2.1525618086425604,
19
+ 0.7003541659101092
20
+ ],
21
+ "std": [
22
+ 1.6720365948940972,
23
+ 1.3768414767482617,
24
+ 0.0009102921174316669
25
+ ],
26
+ "q01": [
27
+ -0.7605057497562686,
28
+ -5.518671947805291,
29
+ 0.6999764703640262
30
+ ],
31
+ "q99": [
32
+ 6.038879433964306,
33
+ -0.5168299981147954,
34
+ 0.7029182461269848
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.24782371371751144,
54
+ 0.5971093380149053
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6642559499983499,
60
+ 0.3751891785184018
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -1.0,
66
+ 4.194193576416142e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 1.0,
72
+ 1.0
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8848381638526917,
78
+ 0.8569459915161133,
79
+ 1.053884744644165
80
+ ],
81
+ "min": [
82
+ -0.5828549861907959,
83
+ -0.8486236333847046,
84
+ -0.3446122705936432
85
+ ],
86
+ "mean": [
87
+ 0.25006971579930243,
88
+ -0.005716176032514759,
89
+ 0.4426027721228362
90
+ ],
91
+ "std": [
92
+ 0.15838384751435522,
93
+ 0.25141504139800513,
94
+ 0.20646816234620438
95
+ ],
96
+ "q01": [
97
+ -0.017003721077380903,
98
+ -0.47738320165276543,
99
+ 0.1917563409753314
100
+ ],
101
+ "q99": [
102
+ 0.5401029199137674,
103
+ 0.4706646609268599,
104
+ 0.6965586537529516
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999949932098389,
110
+ 0.9997354745864868,
111
+ 0.9900747537612915,
112
+ 0.9735639095306396
113
+ ],
114
+ "min": [
115
+ -0.9999914169311523,
116
+ -0.9999884366989136,
117
+ -0.9975369572639465,
118
+ 6.754255821306288e-08
119
+ ],
120
+ "mean": [
121
+ -0.22531628862537245,
122
+ 0.045669797538445966,
123
+ -0.06098055586474368,
124
+ 0.12292005493282424
125
+ ],
126
+ "std": [
127
+ 0.8506341437448487,
128
+ 0.2880763358322842,
129
+ 0.30816573781739026,
130
+ 0.16366165154486362
131
+ ],
132
+ "q01": [
133
+ -0.9957510593732032,
134
+ -0.5584741016891311,
135
+ -0.5371647381159943,
136
+ 0.0013784815838340492
137
+ ],
138
+ "q99": [
139
+ 0.9899375209218226,
140
+ 0.6327878884078038,
141
+ 0.5071539024531994,
142
+ 0.43555975763914345
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05289580300450325,
148
+ 0.00937709677964449
149
+ ],
150
+ "min": [
151
+ -0.011669117026031017,
152
+ -0.06307756155729294
153
+ ],
154
+ "mean": [
155
+ 0.03196506153371015,
156
+ -0.03229791582989581
157
+ ],
158
+ "std": [
159
+ 0.01171640503534049,
160
+ 0.01140016755425199
161
+ ],
162
+ "q01": [
163
+ 0.006290779996034005,
164
+ -0.04059717954551848
165
+ ],
166
+ "q99": [
167
+ 0.04038165838477291,
168
+ -0.006363076424932449
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.006112990930475949,
188
+ 0.002504955933349057,
189
+ -0.001639409958610569,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11478077265783658,
194
+ 0.1002662368589144,
195
+ 0.0807258815139021,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.16969970921613756,
200
+ -0.19919537649977875,
201
+ -0.15012139866371593,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.2466625453958827,
206
+ 0.25040842772679267,
207
+ 0.12777701094549537,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9253067479599785
220
+ ],
221
+ "std": [
222
+ 0.3792397014797737
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.1974093068624371
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.006754222051360652,
244
+ 0.0027854853184013867,
245
+ -0.08087572130330477
246
+ ],
247
+ "std": [
248
+ 0.45779905420789546,
249
+ 0.40578975219522756,
250
+ 0.4062694098132682
251
+ ],
252
+ "q01": [
253
+ -0.968295960290036,
254
+ -0.9240402337368848,
255
+ -0.8249359802202183
256
+ ],
257
+ "q99": [
258
+ 0.8046634915851651,
259
+ 0.9212243025654695,
260
+ 0.838433711437593
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.008633749251857347,
276
+ -0.020406457524911718,
277
+ -0.0027639446362883006
278
+ ],
279
+ "std": [
280
+ 0.11393972872407104,
281
+ 0.13300186268009506,
282
+ 0.12192288302920105
283
+ ],
284
+ "q01": [
285
+ -0.2668579353776187,
286
+ -0.40299251206771697,
287
+ -0.34712957892172586
288
+ ],
289
+ "q99": [
290
+ 0.32748281264383927,
291
+ 0.3179028969292779,
292
+ 0.33459405510046736
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3091635769453286
304
+ ],
305
+ "std": [
306
+ 0.9510122432876397
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7643545615216344
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-3000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/experiment_cfg/metadata.json ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "statistics": {
4
+ "state": {
5
+ "base_position": {
6
+ "max": [
7
+ 7.778976917266846,
8
+ 0.33403685688972473,
9
+ 0.7223060727119446
10
+ ],
11
+ "min": [
12
+ -4.821934223175049,
13
+ -6.785110950469971,
14
+ 0.6994727253913879
15
+ ],
16
+ "mean": [
17
+ 2.577927505176127,
18
+ -2.1525618086425604,
19
+ 0.7003541659101092
20
+ ],
21
+ "std": [
22
+ 1.6720365948940972,
23
+ 1.3768414767482617,
24
+ 0.0009102921174316669
25
+ ],
26
+ "q01": [
27
+ -0.7605057497562686,
28
+ -5.518671947805291,
29
+ 0.6999764703640262
30
+ ],
31
+ "q99": [
32
+ 6.038879433964306,
33
+ -0.5168299981147954,
34
+ 0.7029182461269848
35
+ ]
36
+ },
37
+ "base_rotation": {
38
+ "max": [
39
+ 0.0,
40
+ 0.0,
41
+ 1.0,
42
+ 1.0
43
+ ],
44
+ "min": [
45
+ 0.0,
46
+ 0.0,
47
+ -1.0,
48
+ 0.0
49
+ ],
50
+ "mean": [
51
+ 0.0,
52
+ 0.0,
53
+ 0.24782371371751144,
54
+ 0.5971093380149053
55
+ ],
56
+ "std": [
57
+ 0.0,
58
+ 0.0,
59
+ 0.6642559499983499,
60
+ 0.3751891785184018
61
+ ],
62
+ "q01": [
63
+ 0.0,
64
+ 0.0,
65
+ -1.0,
66
+ 4.194193576416142e-06
67
+ ],
68
+ "q99": [
69
+ 0.0,
70
+ 0.0,
71
+ 1.0,
72
+ 1.0
73
+ ]
74
+ },
75
+ "end_effector_position_relative": {
76
+ "max": [
77
+ 0.8848381638526917,
78
+ 0.8569459915161133,
79
+ 1.053884744644165
80
+ ],
81
+ "min": [
82
+ -0.5828549861907959,
83
+ -0.8486236333847046,
84
+ -0.3446122705936432
85
+ ],
86
+ "mean": [
87
+ 0.25006971579930243,
88
+ -0.005716176032514759,
89
+ 0.4426027721228362
90
+ ],
91
+ "std": [
92
+ 0.15838384751435522,
93
+ 0.25141504139800513,
94
+ 0.20646816234620438
95
+ ],
96
+ "q01": [
97
+ -0.017003721077380903,
98
+ -0.47738320165276543,
99
+ 0.1917563409753314
100
+ ],
101
+ "q99": [
102
+ 0.5401029199137674,
103
+ 0.4706646609268599,
104
+ 0.6965586537529516
105
+ ]
106
+ },
107
+ "end_effector_rotation_relative": {
108
+ "max": [
109
+ 0.9999949932098389,
110
+ 0.9997354745864868,
111
+ 0.9900747537612915,
112
+ 0.9735639095306396
113
+ ],
114
+ "min": [
115
+ -0.9999914169311523,
116
+ -0.9999884366989136,
117
+ -0.9975369572639465,
118
+ 6.754255821306288e-08
119
+ ],
120
+ "mean": [
121
+ -0.22531628862537245,
122
+ 0.045669797538445966,
123
+ -0.06098055586474368,
124
+ 0.12292005493282424
125
+ ],
126
+ "std": [
127
+ 0.8506341437448487,
128
+ 0.2880763358322842,
129
+ 0.30816573781739026,
130
+ 0.16366165154486362
131
+ ],
132
+ "q01": [
133
+ -0.9957510593732032,
134
+ -0.5584741016891311,
135
+ -0.5371647381159943,
136
+ 0.0013784815838340492
137
+ ],
138
+ "q99": [
139
+ 0.9899375209218226,
140
+ 0.6327878884078038,
141
+ 0.5071539024531994,
142
+ 0.43555975763914345
143
+ ]
144
+ },
145
+ "gripper_qpos": {
146
+ "max": [
147
+ 0.05289580300450325,
148
+ 0.00937709677964449
149
+ ],
150
+ "min": [
151
+ -0.011669117026031017,
152
+ -0.06307756155729294
153
+ ],
154
+ "mean": [
155
+ 0.03196506153371015,
156
+ -0.03229791582989581
157
+ ],
158
+ "std": [
159
+ 0.01171640503534049,
160
+ 0.01140016755425199
161
+ ],
162
+ "q01": [
163
+ 0.006290779996034005,
164
+ -0.04059717954551848
165
+ ],
166
+ "q99": [
167
+ 0.04038165838477291,
168
+ -0.006363076424932449
169
+ ]
170
+ }
171
+ },
172
+ "action": {
173
+ "base_motion": {
174
+ "max": [
175
+ 1.0,
176
+ 1.0,
177
+ 1.0,
178
+ 0.0
179
+ ],
180
+ "min": [
181
+ -1.0,
182
+ -1.0,
183
+ -1.0,
184
+ 0.0
185
+ ],
186
+ "mean": [
187
+ 0.006112990930475949,
188
+ 0.002504955933349057,
189
+ -0.001639409958610569,
190
+ 0.0
191
+ ],
192
+ "std": [
193
+ 0.11478077265783658,
194
+ 0.1002662368589144,
195
+ 0.0807258815139021,
196
+ 0.0
197
+ ],
198
+ "q01": [
199
+ -0.16969970921613756,
200
+ -0.19919537649977875,
201
+ -0.15012139866371593,
202
+ 0.0
203
+ ],
204
+ "q99": [
205
+ 0.2466625453958827,
206
+ 0.25040842772679267,
207
+ 0.12777701094549537,
208
+ 0.0
209
+ ]
210
+ },
211
+ "control_mode": {
212
+ "max": [
213
+ 1.0
214
+ ],
215
+ "min": [
216
+ -1.0
217
+ ],
218
+ "mean": [
219
+ -0.9253067479599785
220
+ ],
221
+ "std": [
222
+ 0.3792397014797737
223
+ ],
224
+ "q01": [
225
+ -1.0
226
+ ],
227
+ "q99": [
228
+ -0.1974093068624371
229
+ ]
230
+ },
231
+ "end_effector_position": {
232
+ "max": [
233
+ 1.0,
234
+ 1.0,
235
+ 1.0
236
+ ],
237
+ "min": [
238
+ -1.0,
239
+ -1.0,
240
+ -1.0
241
+ ],
242
+ "mean": [
243
+ -0.006754222051360652,
244
+ 0.0027854853184013867,
245
+ -0.08087572130330477
246
+ ],
247
+ "std": [
248
+ 0.45779905420789546,
249
+ 0.40578975219522756,
250
+ 0.4062694098132682
251
+ ],
252
+ "q01": [
253
+ -0.968295960290036,
254
+ -0.9240402337368848,
255
+ -0.8249359802202183
256
+ ],
257
+ "q99": [
258
+ 0.8046634915851651,
259
+ 0.9212243025654695,
260
+ 0.838433711437593
261
+ ]
262
+ },
263
+ "end_effector_rotation": {
264
+ "max": [
265
+ 1.0,
266
+ 1.0,
267
+ 1.0
268
+ ],
269
+ "min": [
270
+ -1.0,
271
+ -1.0,
272
+ -1.0
273
+ ],
274
+ "mean": [
275
+ 0.008633749251857347,
276
+ -0.020406457524911718,
277
+ -0.0027639446362883006
278
+ ],
279
+ "std": [
280
+ 0.11393972872407104,
281
+ 0.13300186268009506,
282
+ 0.12192288302920105
283
+ ],
284
+ "q01": [
285
+ -0.2668579353776187,
286
+ -0.40299251206771697,
287
+ -0.34712957892172586
288
+ ],
289
+ "q99": [
290
+ 0.32748281264383927,
291
+ 0.3179028969292779,
292
+ 0.33459405510046736
293
+ ]
294
+ },
295
+ "gripper_close": {
296
+ "max": [
297
+ 1.0
298
+ ],
299
+ "min": [
300
+ -1.0
301
+ ],
302
+ "mean": [
303
+ -0.3091635769453286
304
+ ],
305
+ "std": [
306
+ 0.9510122432876397
307
+ ],
308
+ "q01": [
309
+ -1.0
310
+ ],
311
+ "q99": [
312
+ 0.7643545615216344
313
+ ]
314
+ }
315
+ }
316
+ },
317
+ "modalities": {
318
+ "video": {
319
+ "robot0_eye_in_hand": {
320
+ "resolution": [
321
+ 256,
322
+ 256
323
+ ],
324
+ "channels": 3,
325
+ "fps": 20.0
326
+ },
327
+ "robot0_agentview_left": {
328
+ "resolution": [
329
+ 256,
330
+ 256
331
+ ],
332
+ "channels": 3,
333
+ "fps": 20.0
334
+ },
335
+ "robot0_agentview_right": {
336
+ "resolution": [
337
+ 256,
338
+ 256
339
+ ],
340
+ "channels": 3,
341
+ "fps": 20.0
342
+ }
343
+ },
344
+ "state": {
345
+ "base_position": {
346
+ "absolute": true,
347
+ "rotation_type": null,
348
+ "shape": [
349
+ 3
350
+ ],
351
+ "continuous": true
352
+ },
353
+ "base_rotation": {
354
+ "absolute": true,
355
+ "rotation_type": "quaternion",
356
+ "shape": [
357
+ 4
358
+ ],
359
+ "continuous": true
360
+ },
361
+ "end_effector_position_relative": {
362
+ "absolute": true,
363
+ "rotation_type": null,
364
+ "shape": [
365
+ 3
366
+ ],
367
+ "continuous": true
368
+ },
369
+ "end_effector_rotation_relative": {
370
+ "absolute": true,
371
+ "rotation_type": "quaternion",
372
+ "shape": [
373
+ 4
374
+ ],
375
+ "continuous": true
376
+ },
377
+ "gripper_qpos": {
378
+ "absolute": true,
379
+ "rotation_type": null,
380
+ "shape": [
381
+ 2
382
+ ],
383
+ "continuous": true
384
+ }
385
+ },
386
+ "action": {
387
+ "base_motion": {
388
+ "absolute": true,
389
+ "rotation_type": null,
390
+ "shape": [
391
+ 4
392
+ ],
393
+ "continuous": true
394
+ },
395
+ "control_mode": {
396
+ "absolute": true,
397
+ "rotation_type": null,
398
+ "shape": [
399
+ 1
400
+ ],
401
+ "continuous": true
402
+ },
403
+ "end_effector_position": {
404
+ "absolute": true,
405
+ "rotation_type": null,
406
+ "shape": [
407
+ 3
408
+ ],
409
+ "continuous": true
410
+ },
411
+ "end_effector_rotation": {
412
+ "absolute": true,
413
+ "rotation_type": "axis_angle",
414
+ "shape": [
415
+ 3
416
+ ],
417
+ "continuous": true
418
+ },
419
+ "gripper_close": {
420
+ "absolute": true,
421
+ "rotation_type": null,
422
+ "shape": [
423
+ 1
424
+ ],
425
+ "continuous": true
426
+ }
427
+ }
428
+ },
429
+ "embodiment_tag": "new_embodiment"
430
+ }
431
+ }