Hugging Face's logo

seungkukim
/
dit4dit_gr1_posttrain_attn_align_2pair-100k

Safetensors
WAM_DIT4DIT
Model card Files
xet
 Community
seungkukim commited on
Commit
886c749
·
verified ·
1 Parent(s): d691908

Upload folder using huggingface_hub

Browse files
Files changed (14) hide show
  1. config.json +180 -0
  2. embodiment_id.json +39 -0
  3. experiment_cfg/conf.yaml +431 -0
  4. experiment_cfg/config.yaml +498 -0
  5. experiment_cfg/dataset_statistics.json +497 -0
  6. experiment_cfg/final_model_config.json +102 -0
  7. model-00001-of-00005.safetensors +3 -0
  8. model-00002-of-00005.safetensors +3 -0
  9. model-00003-of-00005.safetensors +3 -0
  10. model-00004-of-00005.safetensors +3 -0
  11. model-00005-of-00005.safetensors +3 -0
  12. model.safetensors.index.json +0 -0
  13. processor_config.json +60 -0
  14. statistics.json +497 -0
config.json ADDED
@@ -0,0 +1,180 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "",
3
+ "action_dit_size": "DiT-B",
4
+ "action_head_model_type": "dit",
5
+ "action_horizon": 16,
6
+ "action_loss_weight": 1.0,
7
+ "action_text_kv": false,
8
+ "add_cross_attention": false,
9
+ "add_pos_embed": true,
10
+ "apply_sincos_state_encoding": true,
11
+ "architectures": [
12
+ "WAMDiT4DiT"
13
+ ],
14
+ "attn_align": true,
15
+ "attn_align_frame_offsets": [
16
+ 0,
17
+ 6,
18
+ 14
19
+ ],
20
+ "attn_align_keep_heads": 12,
21
+ "attn_align_lam_ckpt": "",
22
+ "attn_align_stop_step": 60000,
23
+ "attn_align_student_blocks": [
24
+ 6,
25
+ 8
26
+ ],
27
+ "attn_align_teacher_layers": [
28
+ 22,
29
+ 23
30
+ ],
31
+ "attn_align_weight": 0.5,
32
+ "attn_dropout": 0.2,
33
+ "backbone_embedding_dim": 2048,
34
+ "backbone_trainable_params_fp32": true,
35
+ "bad_words_ids": null,
36
+ "begin_suppress_tokens": null,
37
+ "bos_token_id": null,
38
+ "chunk_size_feed_forward": 0,
39
+ "color_jitter_params": null,
40
+ "conditional_frame_timestep": 0.0001,
41
+ "cosmos_model_path": "nvidia/Cosmos-Predict2.5-2B",
42
+ "cosmos_model_revision": "diffusers/base/post-trained",
43
+ "crop_fraction": 0.95,
44
+ "cross_attention_hidden_size": null,
45
+ "decoder_start_token_id": null,
46
+ "diffusion_model_cfg": {
47
+ "attention_head_dim": 64,
48
+ "dropout": 0.2,
49
+ "final_dropout": true,
50
+ "interleave_self_attention": true,
51
+ "norm_type": "ada_norm",
52
+ "num_attention_heads": 12,
53
+ "num_layers": 16,
54
+ "output_dim": 2560,
55
+ "positional_embeddings": null
56
+ },
57
+ "diversity_penalty": 0.0,
58
+ "do_sample": false,
59
+ "dtype": "bfloat16",
60
+ "early_stopping": false,
61
+ "encoder_no_repeat_ngram_size": 0,
62
+ "eos_token_id": null,
63
+ "exponential_decay_length_penalty": null,
64
+ "feature_extraction_aggregate_all": false,
65
+ "feature_extraction_aggregate_layer_window": 0,
66
+ "feature_extraction_layer": 17,
67
+ "finetuning_task": null,
68
+ "fm_high_sigma_min": 0.98,
69
+ "fm_high_sigma_ratio": null,
70
+ "fm_loss_weight": 1.0,
71
+ "fm_time_distribution": "uniform",
72
+ "forced_bos_token_id": null,
73
+ "forced_eos_token_id": null,
74
+ "formalize_language": true,
75
+ "frame_downsample_factor": 2,
76
+ "freeze_modules_pattern": "text_encoder,vae,transformer",
77
+ "freeze_whole_backbone": false,
78
+ "future_loss_type": "flow_matching",
79
+ "future_num_inference_steps": 1,
80
+ "hidden_size": 2560,
81
+ "history_num_clean_frames": 1,
82
+ "history_obs_noise_timestep": 0.0,
83
+ "id2label": {
84
+ "0": "LABEL_0",
85
+ "1": "LABEL_1"
86
+ },
87
+ "image_crop_size": null,
88
+ "image_resize_backend": "cv2_linear",
89
+ "image_target_size": null,
90
+ "input_embedding_dim": 768,
91
+ "is_decoder": false,
92
+ "is_encoder_decoder": false,
93
+ "label2id": {
94
+ "LABEL_0": 0,
95
+ "LABEL_1": 1
96
+ },
97
+ "length_penalty": 1.0,
98
+ "load_bf16": true,
99
+ "max_action_dim": 32,
100
+ "max_length": 20,
101
+ "max_num_embodiments": 36,
102
+ "max_seq_len": 1024,
103
+ "max_state_dim": 64,
104
+ "meta_queries_mode": "meta_only",
105
+ "min_length": 0,
106
+ "model_dtype": "bfloat16",
107
+ "model_type": "WAM_DIT4DIT",
108
+ "n_meta_queries": 8,
109
+ "n_tune_layers_below_feature_extraction": 0,
110
+ "new_embodiment_train_ratio": 0,
111
+ "no_repeat_ngram_size": 0,
112
+ "noise_beta_alpha": 1.5,
113
+ "noise_beta_beta": 1.0,
114
+ "noise_s": 0.999,
115
+ "num_beam_groups": 1,
116
+ "num_beams": 1,
117
+ "num_frames_in": 1,
118
+ "num_frames_out": 17,
119
+ "num_inference_timesteps": 4,
120
+ "num_return_sequences": 1,
121
+ "num_timestep_buckets": 1000,
122
+ "output_attentions": false,
123
+ "output_hidden_states": false,
124
+ "output_scores": false,
125
+ "pad_token_id": null,
126
+ "prefix": null,
127
+ "problem_type": null,
128
+ "pruned_heads": {},
129
+ "random_rotation_angle": null,
130
+ "remove_invalid_values": false,
131
+ "repeated_diffusion_steps": 4,
132
+ "repetition_penalty": 1.0,
133
+ "reproject_vision": false,
134
+ "return_dict": true,
135
+ "return_dict_in_generate": false,
136
+ "selfflow_ema_decay": 0.9999,
137
+ "selfflow_mask_ratio": 0.1,
138
+ "selfflow_rep": false,
139
+ "selfflow_rep_layer": -1,
140
+ "selfflow_rep_weight": 0.8,
141
+ "selfflow_teacher_noise": 0.0,
142
+ "sep_token_id": null,
143
+ "shortest_image_edge": null,
144
+ "state_additive_noise_scale": 0.0,
145
+ "state_dropout_prob": 0.0,
146
+ "suppress_tokens": null,
147
+ "task_specific_params": null,
148
+ "temperature": 1.0,
149
+ "tf_legacy_loss": false,
150
+ "tie_encoder_decoder": false,
151
+ "tie_word_embeddings": true,
152
+ "tokenizer_class": null,
153
+ "top_k": 50,
154
+ "top_p": 1.0,
155
+ "torchscript": false,
156
+ "training_mode": "action",
157
+ "transformers_version": "4.57.0",
158
+ "truncate_above_feature_extraction": true,
159
+ "tune_crossattn_proj": false,
160
+ "tune_diffusion_model": true,
161
+ "tune_layers_above_feature_extraction": true,
162
+ "tune_llm": false,
163
+ "tune_patch_embed": false,
164
+ "tune_projector": true,
165
+ "tune_time_embed": false,
166
+ "tune_top_llm_layers": 4,
167
+ "tune_vgln": true,
168
+ "tune_visual": false,
169
+ "typical_p": 1.0,
170
+ "use_albumentations_transforms": true,
171
+ "use_bfloat16": false,
172
+ "use_meta_queries": false,
173
+ "use_meta_queries_in_fm": false,
174
+ "use_noise_frames_in_extraction": true,
175
+ "use_relative_action": true,
176
+ "use_vgln": false,
177
+ "use_vlm_meta": false,
178
+ "video_train_crop": false,
179
+ "vl_adapter_num_layers": 0
180
+ }
embodiment_id.json ADDED
@@ -0,0 +1,39 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": 0,
3
+ "fractal20220817_data": 1,
4
+ "kuka": 2,
5
+ "bridge_orig": 3,
6
+ "taco_play": 4,
7
+ "jaco_play": 5,
8
+ "berkeley_cable_routing": 6,
9
+ "roboturk": 7,
10
+ "viola": 8,
11
+ "berkeley_autolab_ur5": 9,
12
+ "toto": 10,
13
+ "language_table": 11,
14
+ "stanford_hydra_dataset_converted_externally_to_rlds": 12,
15
+ "austin_buds_dataset_converted_externally_to_rlds": 13,
16
+ "nyu_franka_play_dataset_converted_externally_to_rlds": 14,
17
+ "furniture_bench_dataset_converted_externally_to_rlds": 15,
18
+ "ucsd_kitchen_dataset_converted_externally_to_rlds": 16,
19
+ "austin_sailor_dataset_converted_externally_to_rlds": 17,
20
+ "austin_sirius_dataset_converted_externally_to_rlds": 18,
21
+ "dlr_edan_shared_control_converted_externally_to_rlds": 19,
22
+ "iamlab_cmu_pickup_insert_converted_externally_to_rlds": 20,
23
+ "utaustin_mutex": 21,
24
+ "berkeley_fanuc_manipulation": 22,
25
+ "cmu_stretch": 23,
26
+ "bc_z": 24,
27
+ "fmb_dataset": 25,
28
+ "dobbe": 26,
29
+ "droid": 27,
30
+ "agibot_dexhand": 28,
31
+ "agibot_gripper": 29,
32
+ "galaxea": 30,
33
+ "humanoid_everyday_g1": 31,
34
+ "humanoid_everyday_h1": 32,
35
+ "action_net": 33,
36
+ "neural_gr1": 34,
37
+ "simulation_gr1": 35,
38
+ "libero_panda": 2
39
+ }
experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,431 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ load_config_path: null
2
+ model:
3
+ model_type: WAM_DIT4DIT
4
+ model_dtype: bfloat16
5
+ cosmos_model_path: nvidia/Cosmos-Predict2.5-2B
6
+ cosmos_model_revision: diffusers/base/post-trained
7
+ backbone_embedding_dim: 2048
8
+ num_frames_in: 1
9
+ num_frames_out: 17
10
+ frame_downsample_factor: 2
11
+ feature_extraction_layer: 17
12
+ feature_extraction_aggregate_all: false
13
+ feature_extraction_aggregate_layer_window: 0
14
+ truncate_above_feature_extraction: true
15
+ vl_adapter_num_layers: 0
16
+ tune_layers_above_feature_extraction: true
17
+ n_tune_layers_below_feature_extraction: 0
18
+ tune_patch_embed: false
19
+ tune_time_embed: false
20
+ tune_crossattn_proj: false
21
+ freeze_modules_pattern: text_encoder,vae,transformer
22
+ freeze_whole_backbone: false
23
+ fm_loss_weight: 1.0
24
+ action_loss_weight: 1.0
25
+ conditional_frame_timestep: 0.0001
26
+ history_obs_noise_timestep: 0.0
27
+ history_num_clean_frames: 1
28
+ selfflow_rep: false
29
+ selfflow_rep_layer: -1
30
+ selfflow_mask_ratio: 0.1
31
+ selfflow_ema_decay: 0.9999
32
+ selfflow_rep_weight: 0.8
33
+ selfflow_teacher_noise: 0.0
34
+ attn_align: true
35
+ attn_align_lam_ckpt: ''
36
+ attn_align_student_blocks:
37
+ - 6
38
+ - 8
39
+ attn_align_teacher_layers:
40
+ - 22
41
+ - 23
42
+ attn_align_frame_offsets:
43
+ - 0
44
+ - 6
45
+ - 14
46
+ attn_align_weight: 0.5
47
+ attn_align_keep_heads: 12
48
+ attn_align_stop_step: 60000
49
+ fm_time_distribution: uniform
50
+ fm_high_sigma_ratio: null
51
+ fm_high_sigma_min: 0.98
52
+ future_loss_type: flow_matching
53
+ future_num_inference_steps: 1
54
+ use_noise_frames_in_extraction: true
55
+ use_meta_queries: false
56
+ n_meta_queries: 8
57
+ meta_queries_mode: meta_only
58
+ use_meta_queries_in_fm: false
59
+ use_vlm_meta: false
60
+ use_vgln: false
61
+ tune_vgln: true
62
+ action_text_kv: false
63
+ state_dropout_prob: 0.0
64
+ state_additive_noise_scale: 0.0
65
+ training_mode: action
66
+ action_head_model_type: dit
67
+ max_state_dim: 64
68
+ apply_sincos_state_encoding: true
69
+ video_train_crop: false
70
+ image_resize_backend: cv2_linear
71
+ max_action_dim: 32
72
+ action_horizon: 16
73
+ hidden_size: 2560
74
+ input_embedding_dim: 768
75
+ repeated_diffusion_steps: 4
76
+ add_pos_embed: true
77
+ max_seq_len: 1024
78
+ attn_dropout: 0.2
79
+ action_dit_size: DiT-B
80
+ diffusion_model_cfg:
81
+ num_attention_heads: 12
82
+ attention_head_dim: 64
83
+ num_layers: 16
84
+ output_dim: 2560
85
+ dropout: 0.2
86
+ final_dropout: true
87
+ interleave_self_attention: true
88
+ norm_type: ada_norm
89
+ positional_embeddings: null
90
+ num_inference_timesteps: 4
91
+ noise_beta_alpha: 1.5
92
+ noise_beta_beta: 1.0
93
+ noise_s: 0.999
94
+ num_timestep_buckets: 1000
95
+ tune_projector: true
96
+ tune_diffusion_model: true
97
+ formalize_language: true
98
+ use_albumentations_transforms: true
99
+ crop_fraction: 0.95
100
+ random_rotation_angle: null
101
+ color_jitter_params: null
102
+ image_crop_size: null
103
+ image_target_size: null
104
+ shortest_image_edge: null
105
+ max_num_embodiments: 36
106
+ new_embodiment_train_ratio: 0.0
107
+ data:
108
+ datasets:
109
+ - dataset_paths:
110
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PnPBottleToCabinetClose_GR1ArmsAndWaistFourierHands_1000
111
+ embodiment_tag: new_embodiment
112
+ mix_ratio: 1.0
113
+ dataset_type: physical_embodiment
114
+ val_dataset_path: null
115
+ - dataset_paths:
116
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PnPCanToDrawerClose_GR1ArmsAndWaistFourierHands_1000
117
+ embodiment_tag: new_embodiment
118
+ mix_ratio: 1.0
119
+ dataset_type: physical_embodiment
120
+ val_dataset_path: null
121
+ - dataset_paths:
122
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PnPCupToDrawerClose_GR1ArmsAndWaistFourierHands_1000
123
+ embodiment_tag: new_embodiment
124
+ mix_ratio: 1.0
125
+ dataset_type: physical_embodiment
126
+ val_dataset_path: null
127
+ - dataset_paths:
128
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PnPMilkToMicrowaveClose_GR1ArmsAndWaistFourierHands_1000
129
+ embodiment_tag: new_embodiment
130
+ mix_ratio: 1.0
131
+ dataset_type: physical_embodiment
132
+ val_dataset_path: null
133
+ - dataset_paths:
134
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PnPPotatoToMicrowaveClose_GR1ArmsAndWaistFourierHands_1000
135
+ embodiment_tag: new_embodiment
136
+ mix_ratio: 1.0
137
+ dataset_type: physical_embodiment
138
+ val_dataset_path: null
139
+ - dataset_paths:
140
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PnPWineToCabinetClose_GR1ArmsAndWaistFourierHands_1000
141
+ embodiment_tag: new_embodiment
142
+ mix_ratio: 1.0
143
+ dataset_type: physical_embodiment
144
+ val_dataset_path: null
145
+ - dataset_paths:
146
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromCuttingboardToBasketSplitA_GR1ArmsAndWaistFourierHands_1000
147
+ embodiment_tag: new_embodiment
148
+ mix_ratio: 1.0
149
+ dataset_type: physical_embodiment
150
+ val_dataset_path: null
151
+ - dataset_paths:
152
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromCuttingboardToCardboardboxSplitA_GR1ArmsAndWaistFourierHands_1000
153
+ embodiment_tag: new_embodiment
154
+ mix_ratio: 1.0
155
+ dataset_type: physical_embodiment
156
+ val_dataset_path: null
157
+ - dataset_paths:
158
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromCuttingboardToPanSplitA_GR1ArmsAndWaistFourierHands_1000
159
+ embodiment_tag: new_embodiment
160
+ mix_ratio: 1.0
161
+ dataset_type: physical_embodiment
162
+ val_dataset_path: null
163
+ - dataset_paths:
164
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromCuttingboardToPotSplitA_GR1ArmsAndWaistFourierHands_1000
165
+ embodiment_tag: new_embodiment
166
+ mix_ratio: 1.0
167
+ dataset_type: physical_embodiment
168
+ val_dataset_path: null
169
+ - dataset_paths:
170
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromCuttingboardToTieredbasketSplitA_GR1ArmsAndWaistFourierHands_1000
171
+ embodiment_tag: new_embodiment
172
+ mix_ratio: 1.0
173
+ dataset_type: physical_embodiment
174
+ val_dataset_path: null
175
+ - dataset_paths:
176
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlacematToBasketSplitA_GR1ArmsAndWaistFourierHands_1000
177
+ embodiment_tag: new_embodiment
178
+ mix_ratio: 1.0
179
+ dataset_type: physical_embodiment
180
+ val_dataset_path: null
181
+ - dataset_paths:
182
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlacematToBowlSplitA_GR1ArmsAndWaistFourierHands_1000
183
+ embodiment_tag: new_embodiment
184
+ mix_ratio: 1.0
185
+ dataset_type: physical_embodiment
186
+ val_dataset_path: null
187
+ - dataset_paths:
188
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlacematToPlateSplitA_GR1ArmsAndWaistFourierHands_1000
189
+ embodiment_tag: new_embodiment
190
+ mix_ratio: 1.0
191
+ dataset_type: physical_embodiment
192
+ val_dataset_path: null
193
+ - dataset_paths:
194
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlacematToTieredshelfSplitA_GR1ArmsAndWaistFourierHands_1000
195
+ embodiment_tag: new_embodiment
196
+ mix_ratio: 1.0
197
+ dataset_type: physical_embodiment
198
+ val_dataset_path: null
199
+ - dataset_paths:
200
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlateToBowlSplitA_GR1ArmsAndWaistFourierHands_1000
201
+ embodiment_tag: new_embodiment
202
+ mix_ratio: 1.0
203
+ dataset_type: physical_embodiment
204
+ val_dataset_path: null
205
+ - dataset_paths:
206
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlateToCardboardboxSplitA_GR1ArmsAndWaistFourierHands_1000
207
+ embodiment_tag: new_embodiment
208
+ mix_ratio: 1.0
209
+ dataset_type: physical_embodiment
210
+ val_dataset_path: null
211
+ - dataset_paths:
212
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlateToPanSplitA_GR1ArmsAndWaistFourierHands_1000
213
+ embodiment_tag: new_embodiment
214
+ mix_ratio: 1.0
215
+ dataset_type: physical_embodiment
216
+ val_dataset_path: null
217
+ - dataset_paths:
218
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlateToPlateSplitA_GR1ArmsAndWaistFourierHands_1000
219
+ embodiment_tag: new_embodiment
220
+ mix_ratio: 1.0
221
+ dataset_type: physical_embodiment
222
+ val_dataset_path: null
223
+ - dataset_paths:
224
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromTrayToCardboardboxSplitA_GR1ArmsAndWaistFourierHands_1000
225
+ embodiment_tag: new_embodiment
226
+ mix_ratio: 1.0
227
+ dataset_type: physical_embodiment
228
+ val_dataset_path: null
229
+ - dataset_paths:
230
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromTrayToPlateSplitA_GR1ArmsAndWaistFourierHands_1000
231
+ embodiment_tag: new_embodiment
232
+ mix_ratio: 1.0
233
+ dataset_type: physical_embodiment
234
+ val_dataset_path: null
235
+ - dataset_paths:
236
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromTrayToPotSplitA_GR1ArmsAndWaistFourierHands_1000
237
+ embodiment_tag: new_embodiment
238
+ mix_ratio: 1.0
239
+ dataset_type: physical_embodiment
240
+ val_dataset_path: null
241
+ - dataset_paths:
242
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromTrayToTieredbasketSplitA_GR1ArmsAndWaistFourierHands_1000
243
+ embodiment_tag: new_embodiment
244
+ mix_ratio: 1.0
245
+ dataset_type: physical_embodiment
246
+ val_dataset_path: null
247
+ - dataset_paths:
248
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromTrayToTieredshelfSplitA_GR1ArmsAndWaistFourierHands_1000
249
+ embodiment_tag: new_embodiment
250
+ mix_ratio: 1.0
251
+ dataset_type: physical_embodiment
252
+ val_dataset_path: null
253
+ modality_configs:
254
+ new_embodiment:
255
+ video:
256
+ delta_indices:
257
+ - 0
258
+ - 2
259
+ - 4
260
+ - 6
261
+ - 8
262
+ - 10
263
+ - 12
264
+ - 14
265
+ - 16
266
+ modality_keys:
267
+ - ego_view
268
+ sin_cos_embedding_keys: null
269
+ mean_std_embedding_keys: null
270
+ action_configs: null
271
+ state:
272
+ delta_indices:
273
+ - 0
274
+ modality_keys:
275
+ - left_arm
276
+ - left_hand
277
+ - right_arm
278
+ - right_hand
279
+ - waist
280
+ sin_cos_embedding_keys:
281
+ - left_arm
282
+ - left_hand
283
+ - right_arm
284
+ - right_hand
285
+ - waist
286
+ mean_std_embedding_keys: null
287
+ action_configs: null
288
+ action:
289
+ delta_indices:
290
+ - 0
291
+ - 1
292
+ - 2
293
+ - 3
294
+ - 4
295
+ - 5
296
+ - 6
297
+ - 7
298
+ - 8
299
+ - 9
300
+ - 10
301
+ - 11
302
+ - 12
303
+ - 13
304
+ - 14
305
+ - 15
306
+ modality_keys:
307
+ - left_arm
308
+ - left_hand
309
+ - right_arm
310
+ - right_hand
311
+ - waist
312
+ sin_cos_embedding_keys: null
313
+ mean_std_embedding_keys: null
314
+ action_configs:
315
+ - rep: ABSOLUTE
316
+ type: NON_EEF
317
+ format: DEFAULT
318
+ state_key: null
319
+ - rep: ABSOLUTE
320
+ type: NON_EEF
321
+ format: DEFAULT
322
+ state_key: null
323
+ - rep: ABSOLUTE
324
+ type: NON_EEF
325
+ format: DEFAULT
326
+ state_key: null
327
+ - rep: ABSOLUTE
328
+ type: NON_EEF
329
+ format: DEFAULT
330
+ state_key: null
331
+ - rep: ABSOLUTE
332
+ type: NON_EEF
333
+ format: DEFAULT
334
+ state_key: null
335
+ language:
336
+ delta_indices:
337
+ - 0
338
+ modality_keys:
339
+ - annotation.human.coarse_action
340
+ sin_cos_embedding_keys: null
341
+ mean_std_embedding_keys: null
342
+ action_configs: null
343
+ dataset_mode: sharded
344
+ download_cache: false
345
+ shard_size: 1024
346
+ episode_sampling_rate: 0.1
347
+ num_shards_per_epoch: 100000
348
+ override_pretraining_statistics: false
349
+ mode: single_turn
350
+ random_chop: 0.0
351
+ mock_dataset_mode: false
352
+ shuffle: true
353
+ seed: 42
354
+ multiprocessing_context: fork
355
+ allow_padding: true
356
+ subsample_ratio: 1.0
357
+ image_crop_size:
358
+ - 244
359
+ - 244
360
+ image_target_size:
361
+ - 224
362
+ - 224
363
+ video_backend: torchcodec
364
+ training:
365
+ output_dir: /NHNHOME/data/seungku/projects/wam/ckpt/wam_dit4dit/finetuned/wam_dit4dit_gr1_unified_24task_ditaction_L17_ts0.0001_pd16_ga1_gb64_gpu_j7153
366
+ experiment_name: dit4dit_gr1_posttrain_attn_align_2pair
367
+ max_steps: 100000
368
+ global_batch_size: 64
369
+ batch_size: null
370
+ gradient_accumulation_steps: 1
371
+ learning_rate: 0.0001
372
+ lr_scheduler_type: cosine_with_min_lr
373
+ min_lr: 5.0e-07
374
+ weight_decay: 0.0
375
+ warmup_ratio: 0.0
376
+ warmup_steps: 5000
377
+ max_grad_norm: 1.0
378
+ new_param_warmup_steps: 0
379
+ reset_stage2_lr_scheduler: false
380
+ adam_beta1: 0.9
381
+ adam_beta2: 0.95
382
+ adam_epsilon: 1.0e-08
383
+ lr_per_module_enabled: false
384
+ lr_base: 3.0e-05
385
+ lr_backbone: 1.0e-05
386
+ lr_action_head: 0.0001
387
+ optim: adamw_torch
388
+ start_from_checkpoint: null
389
+ tf32: true
390
+ fp16: false
391
+ bf16: true
392
+ eval_bf16: true
393
+ logging_steps: 10
394
+ save_steps: 10000
395
+ save_total_limit: 10
396
+ save_vl_model: false
397
+ upload_checkpoints: false
398
+ upload_every: 1000
399
+ upload_last_n_checkpoints: 5
400
+ max_concurrent_uploads: 2
401
+ eval_strategy: 'no'
402
+ eval_steps: 500
403
+ eval_set_split_ratio: 0.1
404
+ eval_batch_size: 2
405
+ save_best_eval_metric_name: ''
406
+ save_best_eval_metric_greater_is_better: true
407
+ deepspeed_stage: 2
408
+ gradient_checkpointing: false
409
+ resume_from_checkpoint: false
410
+ transformers_trust_remote_code: true
411
+ transformers_local_files_only: false
412
+ transformers_cache_dir: null
413
+ transformers_access_token: null
414
+ use_ddp: false
415
+ ddp_bucket_cap_mb: 100
416
+ num_gpus: 4
417
+ dataloader_num_workers: 8
418
+ remove_unused_columns: false
419
+ use_wandb: true
420
+ wandb_project: wam-dit4dit-gr1
421
+ enable_profiling: false
422
+ max_retries: 3
423
+ assert_loss_less_than: null
424
+ add_rl_callback: false
425
+ enable_open_loop_eval: false
426
+ open_loop_eval_traj_ids:
427
+ - 0
428
+ open_loop_eval_steps_per_traj: 100
429
+ open_loop_eval_plot_indices: null
430
+ max_steps: 100000
431
+ save_steps: 10000
experiment_cfg/config.yaml ADDED
@@ -0,0 +1,498 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ !!python/object:gr00t.configs.base_config.Config
2
+ data: !!python/object:gr00t.configs.data.data_config.DataConfig
3
+ allow_padding: true
4
+ dataset_mode: sharded
5
+ datasets:
6
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
7
+ dataset_paths:
8
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PnPBottleToCabinetClose_GR1ArmsAndWaistFourierHands_1000
9
+ dataset_type: physical_embodiment
10
+ embodiment_tag: new_embodiment
11
+ mix_ratio: 1.0
12
+ val_dataset_path: null
13
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
14
+ dataset_paths:
15
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PnPCanToDrawerClose_GR1ArmsAndWaistFourierHands_1000
16
+ dataset_type: physical_embodiment
17
+ embodiment_tag: new_embodiment
18
+ mix_ratio: 1.0
19
+ val_dataset_path: null
20
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
21
+ dataset_paths:
22
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PnPCupToDrawerClose_GR1ArmsAndWaistFourierHands_1000
23
+ dataset_type: physical_embodiment
24
+ embodiment_tag: new_embodiment
25
+ mix_ratio: 1.0
26
+ val_dataset_path: null
27
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
28
+ dataset_paths:
29
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PnPMilkToMicrowaveClose_GR1ArmsAndWaistFourierHands_1000
30
+ dataset_type: physical_embodiment
31
+ embodiment_tag: new_embodiment
32
+ mix_ratio: 1.0
33
+ val_dataset_path: null
34
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
35
+ dataset_paths:
36
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PnPPotatoToMicrowaveClose_GR1ArmsAndWaistFourierHands_1000
37
+ dataset_type: physical_embodiment
38
+ embodiment_tag: new_embodiment
39
+ mix_ratio: 1.0
40
+ val_dataset_path: null
41
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
42
+ dataset_paths:
43
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PnPWineToCabinetClose_GR1ArmsAndWaistFourierHands_1000
44
+ dataset_type: physical_embodiment
45
+ embodiment_tag: new_embodiment
46
+ mix_ratio: 1.0
47
+ val_dataset_path: null
48
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
49
+ dataset_paths:
50
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromCuttingboardToBasketSplitA_GR1ArmsAndWaistFourierHands_1000
51
+ dataset_type: physical_embodiment
52
+ embodiment_tag: new_embodiment
53
+ mix_ratio: 1.0
54
+ val_dataset_path: null
55
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
56
+ dataset_paths:
57
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromCuttingboardToCardboardboxSplitA_GR1ArmsAndWaistFourierHands_1000
58
+ dataset_type: physical_embodiment
59
+ embodiment_tag: new_embodiment
60
+ mix_ratio: 1.0
61
+ val_dataset_path: null
62
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
63
+ dataset_paths:
64
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromCuttingboardToPanSplitA_GR1ArmsAndWaistFourierHands_1000
65
+ dataset_type: physical_embodiment
66
+ embodiment_tag: new_embodiment
67
+ mix_ratio: 1.0
68
+ val_dataset_path: null
69
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
70
+ dataset_paths:
71
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromCuttingboardToPotSplitA_GR1ArmsAndWaistFourierHands_1000
72
+ dataset_type: physical_embodiment
73
+ embodiment_tag: new_embodiment
74
+ mix_ratio: 1.0
75
+ val_dataset_path: null
76
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
77
+ dataset_paths:
78
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromCuttingboardToTieredbasketSplitA_GR1ArmsAndWaistFourierHands_1000
79
+ dataset_type: physical_embodiment
80
+ embodiment_tag: new_embodiment
81
+ mix_ratio: 1.0
82
+ val_dataset_path: null
83
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
84
+ dataset_paths:
85
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlacematToBasketSplitA_GR1ArmsAndWaistFourierHands_1000
86
+ dataset_type: physical_embodiment
87
+ embodiment_tag: new_embodiment
88
+ mix_ratio: 1.0
89
+ val_dataset_path: null
90
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
91
+ dataset_paths:
92
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlacematToBowlSplitA_GR1ArmsAndWaistFourierHands_1000
93
+ dataset_type: physical_embodiment
94
+ embodiment_tag: new_embodiment
95
+ mix_ratio: 1.0
96
+ val_dataset_path: null
97
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
98
+ dataset_paths:
99
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlacematToPlateSplitA_GR1ArmsAndWaistFourierHands_1000
100
+ dataset_type: physical_embodiment
101
+ embodiment_tag: new_embodiment
102
+ mix_ratio: 1.0
103
+ val_dataset_path: null
104
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
105
+ dataset_paths:
106
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlacematToTieredshelfSplitA_GR1ArmsAndWaistFourierHands_1000
107
+ dataset_type: physical_embodiment
108
+ embodiment_tag: new_embodiment
109
+ mix_ratio: 1.0
110
+ val_dataset_path: null
111
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
112
+ dataset_paths:
113
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlateToBowlSplitA_GR1ArmsAndWaistFourierHands_1000
114
+ dataset_type: physical_embodiment
115
+ embodiment_tag: new_embodiment
116
+ mix_ratio: 1.0
117
+ val_dataset_path: null
118
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
119
+ dataset_paths:
120
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlateToCardboardboxSplitA_GR1ArmsAndWaistFourierHands_1000
121
+ dataset_type: physical_embodiment
122
+ embodiment_tag: new_embodiment
123
+ mix_ratio: 1.0
124
+ val_dataset_path: null
125
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
126
+ dataset_paths:
127
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlateToPanSplitA_GR1ArmsAndWaistFourierHands_1000
128
+ dataset_type: physical_embodiment
129
+ embodiment_tag: new_embodiment
130
+ mix_ratio: 1.0
131
+ val_dataset_path: null
132
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
133
+ dataset_paths:
134
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromPlateToPlateSplitA_GR1ArmsAndWaistFourierHands_1000
135
+ dataset_type: physical_embodiment
136
+ embodiment_tag: new_embodiment
137
+ mix_ratio: 1.0
138
+ val_dataset_path: null
139
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
140
+ dataset_paths:
141
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromTrayToCardboardboxSplitA_GR1ArmsAndWaistFourierHands_1000
142
+ dataset_type: physical_embodiment
143
+ embodiment_tag: new_embodiment
144
+ mix_ratio: 1.0
145
+ val_dataset_path: null
146
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
147
+ dataset_paths:
148
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromTrayToPlateSplitA_GR1ArmsAndWaistFourierHands_1000
149
+ dataset_type: physical_embodiment
150
+ embodiment_tag: new_embodiment
151
+ mix_ratio: 1.0
152
+ val_dataset_path: null
153
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
154
+ dataset_paths:
155
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromTrayToPotSplitA_GR1ArmsAndWaistFourierHands_1000
156
+ dataset_type: physical_embodiment
157
+ embodiment_tag: new_embodiment
158
+ mix_ratio: 1.0
159
+ val_dataset_path: null
160
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
161
+ dataset_paths:
162
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromTrayToTieredbasketSplitA_GR1ArmsAndWaistFourierHands_1000
163
+ dataset_type: physical_embodiment
164
+ embodiment_tag: new_embodiment
165
+ mix_ratio: 1.0
166
+ val_dataset_path: null
167
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
168
+ dataset_paths:
169
+ - /NHNHOME/data/seungku/dataset/gr00t_unified/gr1_unified.PosttrainPnPNovelFromTrayToTieredshelfSplitA_GR1ArmsAndWaistFourierHands_1000
170
+ dataset_type: physical_embodiment
171
+ embodiment_tag: new_embodiment
172
+ mix_ratio: 1.0
173
+ val_dataset_path: null
174
+ download_cache: false
175
+ episode_sampling_rate: 0.1
176
+ image_crop_size:
177
+ - 244
178
+ - 244
179
+ image_target_size:
180
+ - 224
181
+ - 224
182
+ mock_dataset_mode: false
183
+ modality_configs:
184
+ new_embodiment:
185
+ action: !!python/object:gr00t.data.types.ModalityConfig
186
+ action_configs:
187
+ - !!python/object:gr00t.data.types.ActionConfig
188
+ format: &id001 !!python/object/apply:gr00t.data.types.ActionFormat
189
+ - default
190
+ rep: &id002 !!python/object/apply:gr00t.data.types.ActionRepresentation
191
+ - absolute
192
+ state_key: null
193
+ type: &id003 !!python/object/apply:gr00t.data.types.ActionType
194
+ - non_eef
195
+ - !!python/object:gr00t.data.types.ActionConfig
196
+ format: *id001
197
+ rep: *id002
198
+ state_key: null
199
+ type: *id003
200
+ - !!python/object:gr00t.data.types.ActionConfig
201
+ format: *id001
202
+ rep: *id002
203
+ state_key: null
204
+ type: *id003
205
+ - !!python/object:gr00t.data.types.ActionConfig
206
+ format: *id001
207
+ rep: *id002
208
+ state_key: null
209
+ type: *id003
210
+ - !!python/object:gr00t.data.types.ActionConfig
211
+ format: *id001
212
+ rep: *id002
213
+ state_key: null
214
+ type: *id003
215
+ delta_indices:
216
+ - 0
217
+ - 1
218
+ - 2
219
+ - 3
220
+ - 4
221
+ - 5
222
+ - 6
223
+ - 7
224
+ - 8
225
+ - 9
226
+ - 10
227
+ - 11
228
+ - 12
229
+ - 13
230
+ - 14
231
+ - 15
232
+ mean_std_embedding_keys: null
233
+ modality_keys:
234
+ - left_arm
235
+ - left_hand
236
+ - right_arm
237
+ - right_hand
238
+ - waist
239
+ sin_cos_embedding_keys: null
240
+ language: !!python/object:gr00t.data.types.ModalityConfig
241
+ action_configs: null
242
+ delta_indices:
243
+ - 0
244
+ mean_std_embedding_keys: null
245
+ modality_keys:
246
+ - annotation.human.coarse_action
247
+ sin_cos_embedding_keys: null
248
+ state: !!python/object:gr00t.data.types.ModalityConfig
249
+ action_configs: null
250
+ delta_indices:
251
+ - 0
252
+ mean_std_embedding_keys: null
253
+ modality_keys:
254
+ - left_arm
255
+ - left_hand
256
+ - right_arm
257
+ - right_hand
258
+ - waist
259
+ sin_cos_embedding_keys:
260
+ - left_arm
261
+ - left_hand
262
+ - right_arm
263
+ - right_hand
264
+ - waist
265
+ video: !!python/object:gr00t.data.types.ModalityConfig
266
+ action_configs: null
267
+ delta_indices:
268
+ - 0
269
+ - 2
270
+ - 4
271
+ - 6
272
+ - 8
273
+ - 10
274
+ - 12
275
+ - 14
276
+ - 16
277
+ mean_std_embedding_keys: null
278
+ modality_keys:
279
+ - ego_view
280
+ sin_cos_embedding_keys: null
281
+ mode: single_turn
282
+ multiprocessing_context: fork
283
+ num_shards_per_epoch: 100000
284
+ override_pretraining_statistics: false
285
+ random_chop: 0.0
286
+ seed: 42
287
+ shard_size: 1024
288
+ shuffle: true
289
+ subsample_ratio: 1.0
290
+ video_backend: torchcodec
291
+ load_config_path: null
292
+ model: !!python/object:gr00t.configs.model.wam_dit4dit.WAMDiT4DiTConfig
293
+ _attn_implementation_internal: null
294
+ _commit_hash: null
295
+ _name_or_path: ''
296
+ _output_attentions: false
297
+ _wam_sparse_checkpoint_config: false
298
+ action_dit_size: DiT-B
299
+ action_head_model_type: dit
300
+ action_text_kv: false
301
+ add_cross_attention: false
302
+ apply_sincos_state_encoding: true
303
+ architectures: null
304
+ attn_align: true
305
+ attn_align_frame_offsets: !!python/tuple
306
+ - 0
307
+ - 6
308
+ - 14
309
+ attn_align_keep_heads: 12
310
+ attn_align_lam_ckpt: ''
311
+ attn_align_stop_step: 60000
312
+ attn_align_student_blocks: !!python/tuple
313
+ - 6
314
+ - 8
315
+ attn_align_teacher_layers: !!python/tuple
316
+ - 22
317
+ - 23
318
+ attn_align_weight: 0.5
319
+ backbone_trainable_params_fp32: true
320
+ bad_words_ids: null
321
+ begin_suppress_tokens: null
322
+ bos_token_id: null
323
+ chunk_size_feed_forward: 0
324
+ color_jitter_params: null
325
+ conditional_frame_timestep: 0.0001
326
+ cosmos_model_revision: diffusers/base/post-trained
327
+ cross_attention_hidden_size: null
328
+ decoder_start_token_id: null
329
+ diffusion_model_cfg:
330
+ attention_head_dim: 64
331
+ dropout: 0.2
332
+ final_dropout: true
333
+ interleave_self_attention: true
334
+ norm_type: ada_norm
335
+ num_attention_heads: 12
336
+ num_layers: 16
337
+ output_dim: 2560
338
+ positional_embeddings: null
339
+ diversity_penalty: 0.0
340
+ do_sample: false
341
+ dtype: null
342
+ early_stopping: false
343
+ encoder_no_repeat_ngram_size: 0
344
+ eos_token_id: null
345
+ exponential_decay_length_penalty: null
346
+ feature_extraction_aggregate_all: false
347
+ feature_extraction_aggregate_layer_window: 0
348
+ feature_extraction_layer: 17
349
+ finetuning_task: null
350
+ forced_bos_token_id: null
351
+ forced_eos_token_id: null
352
+ frame_downsample_factor: 2
353
+ freeze_modules_pattern: text_encoder,vae,transformer
354
+ freeze_whole_backbone: false
355
+ history_num_clean_frames: 1
356
+ history_obs_noise_timestep: 0.0
357
+ id2label:
358
+ 0: LABEL_0
359
+ 1: LABEL_1
360
+ image_resize_backend: cv2_linear
361
+ input_embedding_dim: 768
362
+ is_decoder: false
363
+ is_encoder_decoder: false
364
+ label2id:
365
+ LABEL_0: 0
366
+ LABEL_1: 1
367
+ length_penalty: 1.0
368
+ load_bf16: true
369
+ max_length: 20
370
+ min_length: 0
371
+ model_type: WAM_DIT4DIT
372
+ new_embodiment_train_ratio: 0
373
+ no_repeat_ngram_size: 0
374
+ num_beam_groups: 1
375
+ num_beams: 1
376
+ num_frames_in: 1
377
+ num_frames_out: 17
378
+ num_return_sequences: 1
379
+ output_hidden_states: false
380
+ output_scores: false
381
+ pad_token_id: null
382
+ prefix: null
383
+ problem_type: null
384
+ pruned_heads: {}
385
+ random_rotation_angle: null
386
+ remove_invalid_values: false
387
+ repeated_diffusion_steps: 4
388
+ repetition_penalty: 1.0
389
+ reproject_vision: false
390
+ return_dict: true
391
+ return_dict_in_generate: false
392
+ selfflow_ema_decay: 0.9999
393
+ selfflow_mask_ratio: 0.1
394
+ selfflow_rep: false
395
+ selfflow_rep_layer: -1
396
+ selfflow_rep_weight: 0.8
397
+ selfflow_teacher_noise: 0.0
398
+ sep_token_id: null
399
+ state_dropout_prob: 0.0
400
+ suppress_tokens: null
401
+ task_specific_params: null
402
+ temperature: 1.0
403
+ tf_legacy_loss: false
404
+ tie_encoder_decoder: false
405
+ tie_word_embeddings: true
406
+ tokenizer_class: null
407
+ top_k: 50
408
+ top_p: 1.0
409
+ torchscript: false
410
+ training_mode: action
411
+ transformers_version: null
412
+ truncate_above_feature_extraction: true
413
+ tune_crossattn_proj: false
414
+ tune_diffusion_model: true
415
+ tune_layers_above_feature_extraction: true
416
+ tune_llm: false
417
+ tune_patch_embed: false
418
+ tune_projector: true
419
+ tune_time_embed: false
420
+ tune_top_llm_layers: 4
421
+ tune_vgln: true
422
+ tune_visual: false
423
+ typical_p: 1.0
424
+ use_bfloat16: false
425
+ use_meta_queries: false
426
+ use_meta_queries_in_fm: false
427
+ use_noise_frames_in_extraction: true
428
+ use_relative_action: true
429
+ use_vgln: false
430
+ use_vlm_meta: false
431
+ video_train_crop: false
432
+ vl_adapter_num_layers: 0
433
+ training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
434
+ adam_beta1: 0.9
435
+ adam_beta2: 0.95
436
+ adam_epsilon: 1.0e-08
437
+ add_rl_callback: false
438
+ assert_loss_less_than: null
439
+ batch_size: null
440
+ bf16: true
441
+ dataloader_num_workers: 8
442
+ ddp_bucket_cap_mb: 100
443
+ deepspeed_stage: 2
444
+ enable_open_loop_eval: false
445
+ enable_profiling: false
446
+ eval_batch_size: 2
447
+ eval_bf16: true
448
+ eval_set_split_ratio: 0.1
449
+ eval_steps: 500
450
+ eval_strategy: 'no'
451
+ experiment_name: dit4dit_gr1_posttrain_attn_align_2pair
452
+ fp16: false
453
+ global_batch_size: 64
454
+ gradient_accumulation_steps: 1
455
+ gradient_checkpointing: false
456
+ learning_rate: 0.0001
457
+ logging_steps: 10
458
+ lr_action_head: 0.0001
459
+ lr_backbone: 1.0e-05
460
+ lr_base: 3.0e-05
461
+ lr_per_module_enabled: false
462
+ lr_scheduler_type: cosine_with_min_lr
463
+ max_concurrent_uploads: 2
464
+ max_grad_norm: 1.0
465
+ max_retries: 3
466
+ max_steps: 100000
467
+ min_lr: 5.0e-07
468
+ new_param_warmup_steps: 0
469
+ num_gpus: 4
470
+ open_loop_eval_plot_indices: null
471
+ open_loop_eval_steps_per_traj: 100
472
+ open_loop_eval_traj_ids:
473
+ - 0
474
+ optim: adamw_torch
475
+ output_dir: /NHNHOME/data/seungku/projects/wam/ckpt/wam_dit4dit/finetuned/wam_dit4dit_gr1_unified_24task_ditaction_L17_ts0.0001_pd16_ga1_gb64_gpu_j7153
476
+ remove_unused_columns: false
477
+ reset_stage2_lr_scheduler: false
478
+ resume_from_checkpoint: false
479
+ save_best_eval_metric_greater_is_better: true
480
+ save_best_eval_metric_name: ''
481
+ save_steps: 10000
482
+ save_total_limit: 10
483
+ save_vl_model: false
484
+ start_from_checkpoint: null
485
+ tf32: true
486
+ transformers_access_token: null
487
+ transformers_cache_dir: null
488
+ transformers_local_files_only: false
489
+ transformers_trust_remote_code: true
490
+ upload_checkpoints: false
491
+ upload_every: 1000
492
+ upload_last_n_checkpoints: 5
493
+ use_ddp: false
494
+ use_wandb: true
495
+ wandb_project: wam-dit4dit-gr1
496
+ warmup_ratio: 0.0
497
+ warmup_steps: 5000
498
+ weight_decay: 0.0
experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,497 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "left_arm": {
5
+ "min": [
6
+ -1.6789460182189941,
7
+ -0.026101894676685333,
8
+ -1.3480229377746582,
9
+ -2.5160419940948486,
10
+ -1.9940674304962158,
11
+ -1.3795876502990723,
12
+ -1.1958755254745483
13
+ ],
14
+ "max": [
15
+ 1.3502349853515625,
16
+ 1.2633577585220337,
17
+ 1.2589013576507568,
18
+ 0.001734813442453742,
19
+ 2.521491289138794,
20
+ 1.526998519897461,
21
+ 1.496475338935852
22
+ ],
23
+ "mean": [
24
+ -0.16122485262652236,
25
+ 0.24966383228699365,
26
+ -0.09036625494870047,
27
+ -1.4659872402747474,
28
+ 0.14758988075967255,
29
+ 0.09761730915245914,
30
+ 0.01999631943181157
31
+ ],
32
+ "std": [
33
+ 0.3239551184888045,
34
+ 0.1649267593026261,
35
+ 0.22918102481065847,
36
+ 0.3879892409507838,
37
+ 0.2561021378772761,
38
+ 0.24900657630506887,
39
+ 0.2718770913789641
40
+ ],
41
+ "q01": [
42
+ -1.6401696801185608,
43
+ -0.0021117788553237913,
44
+ -1.136602680683136,
45
+ -2.5065788412094117,
46
+ -0.8428557366132736,
47
+ -1.0090796053409576,
48
+ -1.0905668175220489
49
+ ],
50
+ "q99": [
51
+ 1.1068614256382001,
52
+ 1.135889083147049,
53
+ 0.7082531070709233,
54
+ -0.030924458522349596,
55
+ 1.4936031103134155,
56
+ 1.071806162595749,
57
+ 1.0435830056667328
58
+ ]
59
+ },
60
+ "left_hand": {
61
+ "min": [
62
+ -1.4389894008636475,
63
+ -1.8303323984146118,
64
+ -2.4635109901428223,
65
+ -1.7167329788208008,
66
+ -2.218892812728882,
67
+ -1.526924967765808
68
+ ],
69
+ "max": [
70
+ 2.0179455280303955,
71
+ 2.009377956390381,
72
+ 2.6196515560150146,
73
+ 1.8978251218795776,
74
+ 3.2151029109954834,
75
+ 2.7924649715423584
76
+ ],
77
+ "mean": [
78
+ 0.10059643159396123,
79
+ 0.1020327687613947,
80
+ 0.09068115304035017,
81
+ 0.10139854845601802,
82
+ 0.024744001377257505,
83
+ 0.4319268586502101
84
+ ],
85
+ "std": [
86
+ 0.3535206492560486,
87
+ 0.3365244538793486,
88
+ 0.29920181440956567,
89
+ 0.3381295861400242,
90
+ 0.11513307475829572,
91
+ 0.7351638230912396
92
+ ],
93
+ "q01": [
94
+ -0.48410280704498293,
95
+ -0.26667029321193697,
96
+ -0.22677424252033235,
97
+ -0.2616843205690384,
98
+ -0.2044905386865139,
99
+ -0.046128811538219454
100
+ ],
101
+ "q99": [
102
+ 1.508325070142746,
103
+ 1.5029944777488708,
104
+ 1.5275386273860931,
105
+ 1.5626549482345582,
106
+ 1.1534690403938292,
107
+ 1.9877541887760162
108
+ ]
109
+ },
110
+ "right_arm": {
111
+ "min": [
112
+ -2.0664756298065186,
113
+ -2.1021976470947266,
114
+ -2.296651601791382,
115
+ -2.5318210124969482,
116
+ -3.0013694763183594,
117
+ -1.4908946752548218,
118
+ -1.2908861637115479
119
+ ],
120
+ "max": [
121
+ 1.5148204565048218,
122
+ 0.003278259886428714,
123
+ 1.7851011753082275,
124
+ 0.0016116079641506076,
125
+ 3.0015335083007812,
126
+ 1.4080945253372192,
127
+ 1.4516682624816895
128
+ ],
129
+ "mean": [
130
+ -0.3597793262742925,
131
+ -0.3497321410104632,
132
+ 0.10096844471991062,
133
+ -1.4225566883881888,
134
+ 0.4066467555239796,
135
+ 0.07009127957280724,
136
+ 0.19126981108759838
137
+ ],
138
+ "std": [
139
+ 0.48652922666284487,
140
+ 0.28777208660460085,
141
+ 0.3875673326641109,
142
+ 0.6280022930418089,
143
+ 0.5213473030830882,
144
+ 0.38427665747601053,
145
+ 0.5783359418297189
146
+ ],
147
+ "q01": [
148
+ -1.6940901970863342,
149
+ -1.2592608690261842,
150
+ -0.9485023188591003,
151
+ -2.524535834789276,
152
+ -0.9369082593917847,
153
+ -1.2690116608142852,
154
+ -1.070357494354248
155
+ ],
156
+ "q99": [
157
+ 1.1315186619758606,
158
+ 0.0004920228820992634,
159
+ 1.0889833045005788,
160
+ -0.0453558095544572,
161
+ 1.9339820933341976,
162
+ 1.0967729890346527,
163
+ 1.2506092643737792
164
+ ]
165
+ },
166
+ "right_hand": {
167
+ "min": [
168
+ -1.4716511964797974,
169
+ -2.0171985626220703,
170
+ -2.412123203277588,
171
+ -1.189025640487671,
172
+ -0.8325809836387634,
173
+ -0.21484142541885376
174
+ ],
175
+ "max": [
176
+ 2.7859506607055664,
177
+ 2.1664254665374756,
178
+ 3.0131356716156006,
179
+ 2.69866681098938,
180
+ 1.4733597040176392,
181
+ 2.079848289489746
182
+ ],
183
+ "mean": [
184
+ 0.5148559930967166,
185
+ 0.48166331361668807,
186
+ 0.44755979056935746,
187
+ 0.478348320738102,
188
+ 0.07543329064113398,
189
+ 1.6662840098142622
190
+ ],
191
+ "std": [
192
+ 0.5927067305283013,
193
+ 0.5466689922229184,
194
+ 0.5045839716981343,
195
+ 0.5415835858322722,
196
+ 0.16152719492810785,
197
+ 0.2145459320244959
198
+ ],
199
+ "q01": [
200
+ -0.025738642178475858,
201
+ -0.008413693476468325,
202
+ -0.06011074289679527,
203
+ -0.007438082443550229,
204
+ -0.49244091659784317,
205
+ 0.32079766273498533
206
+ ],
207
+ "q99": [
208
+ 1.6979391312599157,
209
+ 1.5753892362117767,
210
+ 2.761149823665619,
211
+ 1.699782818555832,
212
+ 0.898365780711174,
213
+ 1.8441466689109802
214
+ ]
215
+ },
216
+ "waist": {
217
+ "min": [
218
+ -0.5222951769828796,
219
+ -0.42820972204208374,
220
+ -0.39791223406791687
221
+ ],
222
+ "max": [
223
+ 0.937696099281311,
224
+ 0.3457968235015869,
225
+ 0.47687003016471863
226
+ ],
227
+ "mean": [
228
+ 0.005861744905511538,
229
+ 0.0032679109669212863,
230
+ -4.900589652834239e-05
231
+ ],
232
+ "std": [
233
+ 0.05831863966122946,
234
+ 0.01775894444102605,
235
+ 0.007738613054273984
236
+ ],
237
+ "q01": [
238
+ -0.4424374091625214,
239
+ -0.042612580582499504,
240
+ -0.030681067146360874
241
+ ],
242
+ "q99": [
243
+ 0.2744007483124733,
244
+ 0.11413956061005591,
245
+ 0.04111658498644853
246
+ ]
247
+ }
248
+ },
249
+ "action": {
250
+ "left_arm": {
251
+ "min": [
252
+ -1.710237741470337,
253
+ -0.00042632038821466267,
254
+ -1.4035195112228394,
255
+ -2.801373243331909,
256
+ -2.033860445022583,
257
+ -1.4939793348312378,
258
+ -1.3439968824386597
259
+ ],
260
+ "max": [
261
+ 1.4058303833007812,
262
+ 1.288877010345459,
263
+ 1.2944501638412476,
264
+ 3.1500876502832398e-06,
265
+ 2.563594102859497,
266
+ 1.5000286102294922,
267
+ 1.4998135566711426
268
+ ],
269
+ "mean": [
270
+ -0.163161202489088,
271
+ 0.2451432330223422,
272
+ -0.0891637501772493,
273
+ -1.4614050090312956,
274
+ 0.14608537886912623,
275
+ 0.09256105095846578,
276
+ 0.021076124471922718
277
+ ],
278
+ "std": [
279
+ 0.3295375855877872,
280
+ 0.16417262254509757,
281
+ 0.23146453073009823,
282
+ 0.39842306382564957,
283
+ 0.262282522694757,
284
+ 0.25443646729278696,
285
+ 0.2810018872942849
286
+ ],
287
+ "q01": [
288
+ -1.6453754484653473,
289
+ -7.352335887844674e-05,
290
+ -1.1551570868492127,
291
+ -2.7360924339294432,
292
+ -0.8502070605754852,
293
+ -1.0754927098751068,
294
+ -1.2199868965148926
295
+ ],
296
+ "q99": [
297
+ 1.1092832422256476,
298
+ 1.140557736158371,
299
+ 0.7091880106925965,
300
+ -0.01566793117672205,
301
+ 1.4978188574314117,
302
+ 1.1059463322162628,
303
+ 1.0923435854911796
304
+ ]
305
+ },
306
+ "left_hand": {
307
+ "min": [
308
+ -1.5,
309
+ -1.5,
310
+ -1.5,
311
+ -1.5,
312
+ -3.0,
313
+ 0.0
314
+ ],
315
+ "max": [
316
+ 1.5,
317
+ 1.5,
318
+ 1.5,
319
+ 1.5,
320
+ 3.0,
321
+ 3.0
322
+ ],
323
+ "mean": [
324
+ -0.14977236216266948,
325
+ -0.14977236216266948,
326
+ -0.14977236216266948,
327
+ -0.14977236216266948,
328
+ -0.29954472432533896,
329
+ 0.75
330
+ ],
331
+ "std": [
332
+ 0.7349972944755316,
333
+ 0.7349972944755316,
334
+ 0.7349972944755316,
335
+ 0.7349972944755316,
336
+ 1.4699945889510633,
337
+ 1.299038105676658
338
+ ],
339
+ "q01": [
340
+ -1.5,
341
+ -1.5,
342
+ -1.5,
343
+ -1.5,
344
+ -3.0,
345
+ 0.0
346
+ ],
347
+ "q99": [
348
+ 1.5,
349
+ 1.5,
350
+ 1.5,
351
+ 1.5,
352
+ 3.0,
353
+ 3.0
354
+ ]
355
+ },
356
+ "right_arm": {
357
+ "min": [
358
+ -2.1020591259002686,
359
+ -2.2016870975494385,
360
+ -2.3005123138427734,
361
+ -2.848569631576538,
362
+ -3.000068426132202,
363
+ -1.4995447397232056,
364
+ -1.489180326461792
365
+ ],
366
+ "max": [
367
+ 1.564226508140564,
368
+ 0.00015454324602615088,
369
+ 1.7984894514083862,
370
+ 5.365318065742031e-05,
371
+ 3.0000767707824707,
372
+ 1.4954047203063965,
373
+ 1.4975833892822266
374
+ ],
375
+ "mean": [
376
+ -0.36896711634472007,
377
+ -0.3514934852719306,
378
+ 0.09300540465240678,
379
+ -1.4121803939342499,
380
+ 0.4152552008163184,
381
+ 0.07537192405046274,
382
+ 0.2081388249838104
383
+ ],
384
+ "std": [
385
+ 0.49840835424004803,
386
+ 0.2978266793598536,
387
+ 0.40325129654865804,
388
+ 0.6566084540308043,
389
+ 0.5236517016765893,
390
+ 0.40723760206433485,
391
+ 0.6276659193445018
392
+ ],
393
+ "q01": [
394
+ -1.699393036365509,
395
+ -1.3113291025161744,
396
+ -0.9920944452285767,
397
+ -2.7749218344688416,
398
+ -0.9589024305343627,
399
+ -1.3624527990818023,
400
+ -1.177962738275528
401
+ ],
402
+ "q99": [
403
+ 1.1635798215866089,
404
+ -4.211934747217997e-06,
405
+ 1.1038778030872343,
406
+ -0.03212724145501838,
407
+ 1.9471237277984617,
408
+ 1.2183190023899073,
409
+ 1.4874796867370605
410
+ ]
411
+ },
412
+ "right_hand": {
413
+ "min": [
414
+ -1.5,
415
+ -1.5,
416
+ -1.5,
417
+ -1.5,
418
+ -3.0,
419
+ 3.0
420
+ ],
421
+ "max": [
422
+ 1.5,
423
+ 1.5,
424
+ 1.5,
425
+ 1.5,
426
+ 3.0,
427
+ 3.0
428
+ ],
429
+ "mean": [
430
+ -0.4248296525329351,
431
+ -0.4248296525329351,
432
+ -0.4248296525329351,
433
+ -0.4248296525329351,
434
+ -0.8496593050658702,
435
+ 3.0
436
+ ],
437
+ "std": [
438
+ 1.438573732823235,
439
+ 1.438573732823235,
440
+ 1.438573732823235,
441
+ 1.438573732823235,
442
+ 2.87714746564647,
443
+ 0.0
444
+ ],
445
+ "q01": [
446
+ -1.5,
447
+ -1.5,
448
+ -1.5,
449
+ -1.5,
450
+ -3.0,
451
+ 3.0
452
+ ],
453
+ "q99": [
454
+ 1.5,
455
+ 1.5,
456
+ 1.5,
457
+ 1.5,
458
+ 3.0,
459
+ 3.0
460
+ ]
461
+ },
462
+ "waist": {
463
+ "min": [
464
+ -0.5332962870597839,
465
+ -0.4947744905948639,
466
+ -0.4128894805908203
467
+ ],
468
+ "max": [
469
+ 0.9750414490699768,
470
+ 0.34772318601608276,
471
+ 0.5310259461402893
472
+ ],
473
+ "mean": [
474
+ 0.005974193549870203,
475
+ 0.005230103122206251,
476
+ -0.0002079709078088854
477
+ ],
478
+ "std": [
479
+ 0.06158721324753064,
480
+ 0.018468748717223805,
481
+ 0.005717544758609999
482
+ ],
483
+ "q01": [
484
+ -0.45793325632810594,
485
+ -0.04522607177495956,
486
+ -0.029049276374280453
487
+ ],
488
+ "q99": [
489
+ 0.2767475765943526,
490
+ 0.11469113424420346,
491
+ 0.014944081632420426
492
+ ]
493
+ }
494
+ },
495
+ "relative_action": {}
496
+ }
497
+ }
experiment_cfg/final_model_config.json ADDED
@@ -0,0 +1,102 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "WAM_DIT4DIT",
3
+ "model_dtype": "bfloat16",
4
+ "cosmos_model_path": "nvidia/Cosmos-Predict2.5-2B",
5
+ "cosmos_model_revision": "diffusers/base/post-trained",
6
+ "backbone_embedding_dim": 2048,
7
+ "num_frames_in": 1,
8
+ "num_frames_out": 17,
9
+ "frame_downsample_factor": 2,
10
+ "feature_extraction_layer": 17,
11
+ "feature_extraction_aggregate_all": false,
12
+ "feature_extraction_aggregate_layer_window": 0,
13
+ "truncate_above_feature_extraction": true,
14
+ "vl_adapter_num_layers": 0,
15
+ "tune_layers_above_feature_extraction": true,
16
+ "n_tune_layers_below_feature_extraction": 0,
17
+ "tune_patch_embed": false,
18
+ "tune_time_embed": false,
19
+ "tune_crossattn_proj": false,
20
+ "freeze_modules_pattern": "text_encoder,vae,transformer",
21
+ "freeze_whole_backbone": false,
22
+ "fm_loss_weight": 1.0,
23
+ "action_loss_weight": 1.0,
24
+ "conditional_frame_timestep": 0.0001,
25
+ "history_obs_noise_timestep": 0.0,
26
+ "history_num_clean_frames": 1,
27
+ "selfflow_rep": false,
28
+ "selfflow_rep_layer": -1,
29
+ "selfflow_mask_ratio": 0.1,
30
+ "selfflow_ema_decay": 0.9999,
31
+ "selfflow_rep_weight": 0.8,
32
+ "selfflow_teacher_noise": 0.0,
33
+ "attn_align": true,
34
+ "attn_align_lam_ckpt": "",
35
+ "attn_align_student_blocks": [
36
+ 6,
37
+ 8
38
+ ],
39
+ "attn_align_teacher_layers": [
40
+ 22,
41
+ 23
42
+ ],
43
+ "attn_align_frame_offsets": [
44
+ 0,
45
+ 6,
46
+ 14
47
+ ],
48
+ "attn_align_weight": 0.5,
49
+ "attn_align_keep_heads": 12,
50
+ "attn_align_stop_step": 60000,
51
+ "fm_time_distribution": "uniform",
52
+ "fm_high_sigma_ratio": null,
53
+ "fm_high_sigma_min": 0.98,
54
+ "future_loss_type": "flow_matching",
55
+ "future_num_inference_steps": 1,
56
+ "use_noise_frames_in_extraction": true,
57
+ "use_meta_queries": false,
58
+ "n_meta_queries": 8,
59
+ "meta_queries_mode": "meta_only",
60
+ "use_meta_queries_in_fm": false,
61
+ "use_vlm_meta": false,
62
+ "use_vgln": false,
63
+ "tune_vgln": true,
64
+ "action_text_kv": false,
65
+ "state_dropout_prob": 0.0,
66
+ "state_additive_noise_scale": 0.0,
67
+ "training_mode": "action",
68
+ "action_head_model_type": "dit",
69
+ "max_state_dim": 64,
70
+ "apply_sincos_state_encoding": true,
71
+ "video_train_crop": false,
72
+ "image_resize_backend": "cv2_linear",
73
+ "max_action_dim": 32,
74
+ "action_horizon": 16,
75
+ "hidden_size": 2560,
76
+ "input_embedding_dim": 768,
77
+ "repeated_diffusion_steps": 4,
78
+ "add_pos_embed": true,
79
+ "max_seq_len": 1024,
80
+ "attn_dropout": 0.2,
81
+ "action_dit_size": "DiT-B",
82
+ "diffusion_model_cfg": {
83
+ "num_attention_heads": 12,
84
+ "attention_head_dim": 64,
85
+ "num_layers": 16,
86
+ "output_dim": 2560,
87
+ "dropout": 0.2,
88
+ "final_dropout": true,
89
+ "interleave_self_attention": true,
90
+ "norm_type": "ada_norm",
91
+ "positional_embeddings": null
92
+ },
93
+ "num_inference_timesteps": 4,
94
+ "noise_beta_alpha": 1.5,
95
+ "noise_beta_beta": 1.0,
96
+ "noise_s": 0.999,
97
+ "num_timestep_buckets": 1000,
98
+ "tune_projector": true,
99
+ "tune_diffusion_model": true,
100
+ "max_num_embodiments": 36,
101
+ "new_embodiment_train_ratio": 0
102
+ }
model-00001-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b123925660a6284bdaf8a708ac1ada48173b52726f45a04622b7ef46aa6d2f52
3
+ size 4968261360
model-00002-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f04b9823718d345687dc122cf247d8e60ef2b035fbf59f9fc6fcec87ad81fddf
3
+ size 4991501976
model-00003-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3fed6a0ecb60b3234cd4e8756db4a19b6901c013f9f2747e04c39fc2d10cacc7
3
+ size 4932756776
model-00004-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9359e9f27d9d72a3b1ba599d89599bb0f8f273122c1364a62affa687ffc06332
3
+ size 4991745742
model-00005-of-00005.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e369f095608d317c44107cfe0f34fb3b941986e1e8f082b74c448fe3f74deece
3
+ size 14849672
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
processor_config.json ADDED
@@ -0,0 +1,60 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "processor_class": "WAMDiT4DiTProcessor",
3
+ "downsampled_num_frames_in": 1,
4
+ "downsampled_num_frames_out": 9,
5
+ "frame_downsample_factor": 2,
6
+ "max_state_dim": 64,
7
+ "max_action_dim": 32,
8
+ "max_action_horizon": 16,
9
+ "formalize_language": true,
10
+ "new_embodiment_train_ratio": 0,
11
+ "use_percentiles": false,
12
+ "clip_outliers": true,
13
+ "apply_sincos_state_encoding": true,
14
+ "video_train_crop": false,
15
+ "image_resize_backend": "cv2_linear",
16
+ "use_relative_action": false,
17
+ "embodiment_id_mapping": {
18
+ "new_embodiment": 0,
19
+ "fractal20220817_data": 1,
20
+ "kuka": 2,
21
+ "bridge_orig": 3,
22
+ "taco_play": 4,
23
+ "jaco_play": 5,
24
+ "berkeley_cable_routing": 6,
25
+ "roboturk": 7,
26
+ "viola": 8,
27
+ "berkeley_autolab_ur5": 9,
28
+ "toto": 10,
29
+ "language_table": 11,
30
+ "stanford_hydra_dataset_converted_externally_to_rlds": 12,
31
+ "austin_buds_dataset_converted_externally_to_rlds": 13,
32
+ "nyu_franka_play_dataset_converted_externally_to_rlds": 14,
33
+ "furniture_bench_dataset_converted_externally_to_rlds": 15,
34
+ "ucsd_kitchen_dataset_converted_externally_to_rlds": 16,
35
+ "austin_sailor_dataset_converted_externally_to_rlds": 17,
36
+ "austin_sirius_dataset_converted_externally_to_rlds": 18,
37
+ "dlr_edan_shared_control_converted_externally_to_rlds": 19,
38
+ "iamlab_cmu_pickup_insert_converted_externally_to_rlds": 20,
39
+ "utaustin_mutex": 21,
40
+ "berkeley_fanuc_manipulation": 22,
41
+ "cmu_stretch": 23,
42
+ "bc_z": 24,
43
+ "fmb_dataset": 25,
44
+ "dobbe": 26,
45
+ "droid": 27,
46
+ "agibot_dexhand": 28,
47
+ "agibot_gripper": 29,
48
+ "galaxea": 30,
49
+ "humanoid_everyday_g1": 31,
50
+ "humanoid_everyday_h1": 32,
51
+ "action_net": 33,
52
+ "neural_gr1": 34,
53
+ "simulation_gr1": 35,
54
+ "libero_panda": 2
55
+ },
56
+ "use_vlm_meta": false,
57
+ "vlm_model_name": "huiwon/alinvlm_v1_3",
58
+ "vlm_backbone_model_type": "qwen3_vl",
59
+ "conversation_image_first": false
60
+ }
statistics.json ADDED
@@ -0,0 +1,497 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "left_arm": {
5
+ "min": [
6
+ -1.6789460182189941,
7
+ -0.026101894676685333,
8
+ -1.3480229377746582,
9
+ -2.5160419940948486,
10
+ -1.9940674304962158,
11
+ -1.3795876502990723,
12
+ -1.1958755254745483
13
+ ],
14
+ "max": [
15
+ 1.3502349853515625,
16
+ 1.2633577585220337,
17
+ 1.2589013576507568,
18
+ 0.001734813442453742,
19
+ 2.521491289138794,
20
+ 1.526998519897461,
21
+ 1.496475338935852
22
+ ],
23
+ "mean": [
24
+ -0.16122485262652236,
25
+ 0.24966383228699365,
26
+ -0.09036625494870047,
27
+ -1.4659872402747474,
28
+ 0.14758988075967255,
29
+ 0.09761730915245914,
30
+ 0.01999631943181157
31
+ ],
32
+ "std": [
33
+ 0.3239551184888045,
34
+ 0.1649267593026261,
35
+ 0.22918102481065847,
36
+ 0.3879892409507838,
37
+ 0.2561021378772761,
38
+ 0.24900657630506887,
39
+ 0.2718770913789641
40
+ ],
41
+ "q01": [
42
+ -1.6401696801185608,
43
+ -0.0021117788553237913,
44
+ -1.136602680683136,
45
+ -2.5065788412094117,
46
+ -0.8428557366132736,
47
+ -1.0090796053409576,
48
+ -1.0905668175220489
49
+ ],
50
+ "q99": [
51
+ 1.1068614256382001,
52
+ 1.135889083147049,
53
+ 0.7082531070709233,
54
+ -0.030924458522349596,
55
+ 1.4936031103134155,
56
+ 1.071806162595749,
57
+ 1.0435830056667328
58
+ ]
59
+ },
60
+ "left_hand": {
61
+ "min": [
62
+ -1.4389894008636475,
63
+ -1.8303323984146118,
64
+ -2.4635109901428223,
65
+ -1.7167329788208008,
66
+ -2.218892812728882,
67
+ -1.526924967765808
68
+ ],
69
+ "max": [
70
+ 2.0179455280303955,
71
+ 2.009377956390381,
72
+ 2.6196515560150146,
73
+ 1.8978251218795776,
74
+ 3.2151029109954834,
75
+ 2.7924649715423584
76
+ ],
77
+ "mean": [
78
+ 0.10059643159396123,
79
+ 0.1020327687613947,
80
+ 0.09068115304035017,
81
+ 0.10139854845601802,
82
+ 0.024744001377257505,
83
+ 0.4319268586502101
84
+ ],
85
+ "std": [
86
+ 0.3535206492560486,
87
+ 0.3365244538793486,
88
+ 0.29920181440956567,
89
+ 0.3381295861400242,
90
+ 0.11513307475829572,
91
+ 0.7351638230912396
92
+ ],
93
+ "q01": [
94
+ -0.48410280704498293,
95
+ -0.26667029321193697,
96
+ -0.22677424252033235,
97
+ -0.2616843205690384,
98
+ -0.2044905386865139,
99
+ -0.046128811538219454
100
+ ],
101
+ "q99": [
102
+ 1.508325070142746,
103
+ 1.5029944777488708,
104
+ 1.5275386273860931,
105
+ 1.5626549482345582,
106
+ 1.1534690403938292,
107
+ 1.9877541887760162
108
+ ]
109
+ },
110
+ "right_arm": {
111
+ "min": [
112
+ -2.0664756298065186,
113
+ -2.1021976470947266,
114
+ -2.296651601791382,
115
+ -2.5318210124969482,
116
+ -3.0013694763183594,
117
+ -1.4908946752548218,
118
+ -1.2908861637115479
119
+ ],
120
+ "max": [
121
+ 1.5148204565048218,
122
+ 0.003278259886428714,
123
+ 1.7851011753082275,
124
+ 0.0016116079641506076,
125
+ 3.0015335083007812,
126
+ 1.4080945253372192,
127
+ 1.4516682624816895
128
+ ],
129
+ "mean": [
130
+ -0.3597793262742925,
131
+ -0.3497321410104632,
132
+ 0.10096844471991062,
133
+ -1.4225566883881888,
134
+ 0.4066467555239796,
135
+ 0.07009127957280724,
136
+ 0.19126981108759838
137
+ ],
138
+ "std": [
139
+ 0.48652922666284487,
140
+ 0.28777208660460085,
141
+ 0.3875673326641109,
142
+ 0.6280022930418089,
143
+ 0.5213473030830882,
144
+ 0.38427665747601053,
145
+ 0.5783359418297189
146
+ ],
147
+ "q01": [
148
+ -1.6940901970863342,
149
+ -1.2592608690261842,
150
+ -0.9485023188591003,
151
+ -2.524535834789276,
152
+ -0.9369082593917847,
153
+ -1.2690116608142852,
154
+ -1.070357494354248
155
+ ],
156
+ "q99": [
157
+ 1.1315186619758606,
158
+ 0.0004920228820992634,
159
+ 1.0889833045005788,
160
+ -0.0453558095544572,
161
+ 1.9339820933341976,
162
+ 1.0967729890346527,
163
+ 1.2506092643737792
164
+ ]
165
+ },
166
+ "right_hand": {
167
+ "min": [
168
+ -1.4716511964797974,
169
+ -2.0171985626220703,
170
+ -2.412123203277588,
171
+ -1.189025640487671,
172
+ -0.8325809836387634,
173
+ -0.21484142541885376
174
+ ],
175
+ "max": [
176
+ 2.7859506607055664,
177
+ 2.1664254665374756,
178
+ 3.0131356716156006,
179
+ 2.69866681098938,
180
+ 1.4733597040176392,
181
+ 2.079848289489746
182
+ ],
183
+ "mean": [
184
+ 0.5148559930967166,
185
+ 0.48166331361668807,
186
+ 0.44755979056935746,
187
+ 0.478348320738102,
188
+ 0.07543329064113398,
189
+ 1.6662840098142622
190
+ ],
191
+ "std": [
192
+ 0.5927067305283013,
193
+ 0.5466689922229184,
194
+ 0.5045839716981343,
195
+ 0.5415835858322722,
196
+ 0.16152719492810785,
197
+ 0.2145459320244959
198
+ ],
199
+ "q01": [
200
+ -0.025738642178475858,
201
+ -0.008413693476468325,
202
+ -0.06011074289679527,
203
+ -0.007438082443550229,
204
+ -0.49244091659784317,
205
+ 0.32079766273498533
206
+ ],
207
+ "q99": [
208
+ 1.6979391312599157,
209
+ 1.5753892362117767,
210
+ 2.761149823665619,
211
+ 1.699782818555832,
212
+ 0.898365780711174,
213
+ 1.8441466689109802
214
+ ]
215
+ },
216
+ "waist": {
217
+ "min": [
218
+ -0.5222951769828796,
219
+ -0.42820972204208374,
220
+ -0.39791223406791687
221
+ ],
222
+ "max": [
223
+ 0.937696099281311,
224
+ 0.3457968235015869,
225
+ 0.47687003016471863
226
+ ],
227
+ "mean": [
228
+ 0.005861744905511538,
229
+ 0.0032679109669212863,
230
+ -4.900589652834239e-05
231
+ ],
232
+ "std": [
233
+ 0.05831863966122946,
234
+ 0.01775894444102605,
235
+ 0.007738613054273984
236
+ ],
237
+ "q01": [
238
+ -0.4424374091625214,
239
+ -0.042612580582499504,
240
+ -0.030681067146360874
241
+ ],
242
+ "q99": [
243
+ 0.2744007483124733,
244
+ 0.11413956061005591,
245
+ 0.04111658498644853
246
+ ]
247
+ }
248
+ },
249
+ "action": {
250
+ "left_arm": {
251
+ "min": [
252
+ -1.710237741470337,
253
+ -0.00042632038821466267,
254
+ -1.4035195112228394,
255
+ -2.801373243331909,
256
+ -2.033860445022583,
257
+ -1.4939793348312378,
258
+ -1.3439968824386597
259
+ ],
260
+ "max": [
261
+ 1.4058303833007812,
262
+ 1.288877010345459,
263
+ 1.2944501638412476,
264
+ 3.1500876502832398e-06,
265
+ 2.563594102859497,
266
+ 1.5000286102294922,
267
+ 1.4998135566711426
268
+ ],
269
+ "mean": [
270
+ -0.163161202489088,
271
+ 0.2451432330223422,
272
+ -0.0891637501772493,
273
+ -1.4614050090312956,
274
+ 0.14608537886912623,
275
+ 0.09256105095846578,
276
+ 0.021076124471922718
277
+ ],
278
+ "std": [
279
+ 0.3295375855877872,
280
+ 0.16417262254509757,
281
+ 0.23146453073009823,
282
+ 0.39842306382564957,
283
+ 0.262282522694757,
284
+ 0.25443646729278696,
285
+ 0.2810018872942849
286
+ ],
287
+ "q01": [
288
+ -1.6453754484653473,
289
+ -7.352335887844674e-05,
290
+ -1.1551570868492127,
291
+ -2.7360924339294432,
292
+ -0.8502070605754852,
293
+ -1.0754927098751068,
294
+ -1.2199868965148926
295
+ ],
296
+ "q99": [
297
+ 1.1092832422256476,
298
+ 1.140557736158371,
299
+ 0.7091880106925965,
300
+ -0.01566793117672205,
301
+ 1.4978188574314117,
302
+ 1.1059463322162628,
303
+ 1.0923435854911796
304
+ ]
305
+ },
306
+ "left_hand": {
307
+ "min": [
308
+ -1.5,
309
+ -1.5,
310
+ -1.5,
311
+ -1.5,
312
+ -3.0,
313
+ 0.0
314
+ ],
315
+ "max": [
316
+ 1.5,
317
+ 1.5,
318
+ 1.5,
319
+ 1.5,
320
+ 3.0,
321
+ 3.0
322
+ ],
323
+ "mean": [
324
+ -0.14977236216266948,
325
+ -0.14977236216266948,
326
+ -0.14977236216266948,
327
+ -0.14977236216266948,
328
+ -0.29954472432533896,
329
+ 0.75
330
+ ],
331
+ "std": [
332
+ 0.7349972944755316,
333
+ 0.7349972944755316,
334
+ 0.7349972944755316,
335
+ 0.7349972944755316,
336
+ 1.4699945889510633,
337
+ 1.299038105676658
338
+ ],
339
+ "q01": [
340
+ -1.5,
341
+ -1.5,
342
+ -1.5,
343
+ -1.5,
344
+ -3.0,
345
+ 0.0
346
+ ],
347
+ "q99": [
348
+ 1.5,
349
+ 1.5,
350
+ 1.5,
351
+ 1.5,
352
+ 3.0,
353
+ 3.0
354
+ ]
355
+ },
356
+ "right_arm": {
357
+ "min": [
358
+ -2.1020591259002686,
359
+ -2.2016870975494385,
360
+ -2.3005123138427734,
361
+ -2.848569631576538,
362
+ -3.000068426132202,
363
+ -1.4995447397232056,
364
+ -1.489180326461792
365
+ ],
366
+ "max": [
367
+ 1.564226508140564,
368
+ 0.00015454324602615088,
369
+ 1.7984894514083862,
370
+ 5.365318065742031e-05,
371
+ 3.0000767707824707,
372
+ 1.4954047203063965,
373
+ 1.4975833892822266
374
+ ],
375
+ "mean": [
376
+ -0.36896711634472007,
377
+ -0.3514934852719306,
378
+ 0.09300540465240678,
379
+ -1.4121803939342499,
380
+ 0.4152552008163184,
381
+ 0.07537192405046274,
382
+ 0.2081388249838104
383
+ ],
384
+ "std": [
385
+ 0.49840835424004803,
386
+ 0.2978266793598536,
387
+ 0.40325129654865804,
388
+ 0.6566084540308043,
389
+ 0.5236517016765893,
390
+ 0.40723760206433485,
391
+ 0.6276659193445018
392
+ ],
393
+ "q01": [
394
+ -1.699393036365509,
395
+ -1.3113291025161744,
396
+ -0.9920944452285767,
397
+ -2.7749218344688416,
398
+ -0.9589024305343627,
399
+ -1.3624527990818023,
400
+ -1.177962738275528
401
+ ],
402
+ "q99": [
403
+ 1.1635798215866089,
404
+ -4.211934747217997e-06,
405
+ 1.1038778030872343,
406
+ -0.03212724145501838,
407
+ 1.9471237277984617,
408
+ 1.2183190023899073,
409
+ 1.4874796867370605
410
+ ]
411
+ },
412
+ "right_hand": {
413
+ "min": [
414
+ -1.5,
415
+ -1.5,
416
+ -1.5,
417
+ -1.5,
418
+ -3.0,
419
+ 3.0
420
+ ],
421
+ "max": [
422
+ 1.5,
423
+ 1.5,
424
+ 1.5,
425
+ 1.5,
426
+ 3.0,
427
+ 3.0
428
+ ],
429
+ "mean": [
430
+ -0.4248296525329351,
431
+ -0.4248296525329351,
432
+ -0.4248296525329351,
433
+ -0.4248296525329351,
434
+ -0.8496593050658702,
435
+ 3.0
436
+ ],
437
+ "std": [
438
+ 1.438573732823235,
439
+ 1.438573732823235,
440
+ 1.438573732823235,
441
+ 1.438573732823235,
442
+ 2.87714746564647,
443
+ 0.0
444
+ ],
445
+ "q01": [
446
+ -1.5,
447
+ -1.5,
448
+ -1.5,
449
+ -1.5,
450
+ -3.0,
451
+ 3.0
452
+ ],
453
+ "q99": [
454
+ 1.5,
455
+ 1.5,
456
+ 1.5,
457
+ 1.5,
458
+ 3.0,
459
+ 3.0
460
+ ]
461
+ },
462
+ "waist": {
463
+ "min": [
464
+ -0.5332962870597839,
465
+ -0.4947744905948639,
466
+ -0.4128894805908203
467
+ ],
468
+ "max": [
469
+ 0.9750414490699768,
470
+ 0.34772318601608276,
471
+ 0.5310259461402893
472
+ ],
473
+ "mean": [
474
+ 0.005974193549870203,
475
+ 0.005230103122206251,
476
+ -0.0002079709078088854
477
+ ],
478
+ "std": [
479
+ 0.06158721324753064,
480
+ 0.018468748717223805,
481
+ 0.005717544758609999
482
+ ],
483
+ "q01": [
484
+ -0.45793325632810594,
485
+ -0.04522607177495956,
486
+ -0.029049276374280453
487
+ ],
488
+ "q99": [
489
+ 0.2767475765943526,
490
+ 0.11469113424420346,
491
+ 0.014944081632420426
492
+ ]
493
+ }
494
+ },
495
+ "relative_action": {}
496
+ }
497
+ }