Hugging Face's logo

LiFanxing
/
pack_agx

Safetensors
Model card Files
xet
 Community
LiFanxing commited on
Commit
40a2c24
·
verified ·
1 Parent(s): 5495582

Upload folder using huggingface_hub

Browse files
Files changed (12) hide show
  1. SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/config.json +88 -0
  2. SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/embodiment_id.json +57 -0
  3. SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/experiment_cfg/conf.yaml +248 -0
  4. SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/experiment_cfg/config.yaml +246 -0
  5. SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/experiment_cfg/dataset_statistics.json +817 -0
  6. SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/experiment_cfg/final_model_config.json +56 -0
  7. SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/experiment_cfg/final_processor_config.json +0 -0
  8. SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/model-00001-of-00002.safetensors +3 -0
  9. SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/model-00002-of-00002.safetensors +3 -0
  10. SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/model.safetensors.index.json +0 -0
  11. SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/processor_config.json +1156 -0
  12. SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/statistics.json +0 -0
SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/config.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_horizon": 40,
3
+ "add_pos_embed": true,
4
+ "apply_sincos_state_encoding": false,
5
+ "architectures": [
6
+ "Gr00tN1d7"
7
+ ],
8
+ "attn_dropout": 0.2,
9
+ "attn_implementation": null,
10
+ "backbone_embedding_dim": 2048,
11
+ "backbone_trainable_params_fp32": true,
12
+ "color_jitter_params": {
13
+ "brightness": 0.3,
14
+ "contrast": 0.4,
15
+ "hue": 0.08,
16
+ "saturation": 0.5
17
+ },
18
+ "crop_fraction": 0.95,
19
+ "diffusion_model_cfg": {
20
+ "attention_head_dim": 48,
21
+ "dropout": 0.2,
22
+ "final_dropout": true,
23
+ "interleave_self_attention": true,
24
+ "norm_type": "ada_norm",
25
+ "num_attention_heads": 32,
26
+ "num_layers": 32,
27
+ "output_dim": 1024,
28
+ "positional_embeddings": null
29
+ },
30
+ "dtype": "bfloat16",
31
+ "exclude_state": false,
32
+ "formalize_language": true,
33
+ "hidden_size": 1024,
34
+ "image_crop_size": [
35
+ 230,
36
+ 230
37
+ ],
38
+ "image_target_size": [
39
+ 256,
40
+ 256
41
+ ],
42
+ "letter_box_transform": false,
43
+ "load_bf16": false,
44
+ "max_action_dim": 132,
45
+ "max_num_embodiments": 32,
46
+ "max_seq_len": 1024,
47
+ "max_state_dim": 132,
48
+ "model_dtype": "bfloat16",
49
+ "model_name": "nvidia/Cosmos-Reason2-2B",
50
+ "model_type": "Gr00tN1d7",
51
+ "noise_beta_alpha": 1.5,
52
+ "noise_beta_beta": 1.0,
53
+ "noise_s": 0.999,
54
+ "num_inference_timesteps": 4,
55
+ "num_timestep_buckets": 1000,
56
+ "random_history_crop": true,
57
+ "random_rotation_angle": 0,
58
+ "reproject_vision": false,
59
+ "rtc_ramp_rate": 6.0,
60
+ "select_layer": 16,
61
+ "shortest_image_edge": 256,
62
+ "state_dropout_prob": 0.0,
63
+ "state_gaussian_noise_std": 0.0,
64
+ "transformers_version": "4.57.3",
65
+ "tune_diffusion_model": true,
66
+ "tune_linear": true,
67
+ "tune_llm": false,
68
+ "tune_projector": true,
69
+ "tune_top_llm_layers": 0,
70
+ "tune_visual": true,
71
+ "tune_vlln": true,
72
+ "use_albumentations": true,
73
+ "use_alternate_vl_dit": true,
74
+ "use_flash_attention": true,
75
+ "use_future_tokens": false,
76
+ "use_mean_std": false,
77
+ "use_percentiles": true,
78
+ "use_vl_self_attention": true,
79
+ "use_vlln": true,
80
+ "vl_self_attention_cfg": {
81
+ "attention_head_dim": 64,
82
+ "dropout": 0.2,
83
+ "final_dropout": true,
84
+ "num_attention_heads": 32,
85
+ "num_layers": 4,
86
+ "positional_embeddings": null
87
+ }
88
+ }
SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/embodiment_id.json ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "robocasa_panda_omron": 13,
3
+ "oxe_droid": 17,
4
+ "oxe_fractal": 18,
5
+ "oxe_language_table": 19,
6
+ "oxe_bridge": 20,
7
+ "unknown": 22,
8
+ "gr1_unified": 20,
9
+ "agibot": 26,
10
+ "sim_behavior_r1_pro": 23,
11
+ "xdof": 24,
12
+ "xdof_oss_data": 25,
13
+ "unitree_g1_full_body_with_waist_height_nav_cmd": 25,
14
+ "real_r1_pro_sharpa": 27,
15
+ "real_r1_pro_sharpa_add_view": 27,
16
+ "real_r1_pro_sharpa_relative_arm_joint": 26,
17
+ "real_r1_pro_sharpa_delta_eef": 26,
18
+ "real_r1_pro_sharpa_absolute_eef": 26,
19
+ "real_r1_pro_sharpa_meanstd": 26,
20
+ "real_r1_pro_sharpa_relative_eef": 26,
21
+ "real_r1_pro_sharpa_relative_eef_add_view": 26,
22
+ "real_r1_pro_sharpa_relative_eef_relative_hand": 26,
23
+ "real_r1_pro_sharpa_relative_eef_human": 26,
24
+ "real_r1_pro_sharpa_relative_eef_human_add_view": 26,
25
+ "real_r1_pro_sharpa_relative_eef_human_relative_hand": 26,
26
+ "real_r1_pro_sharpa_relative_eef_egodex": 26,
27
+ "real_r1_pro_sharpa_relative_eef_egodex_relative_hand": 26,
28
+ "real_r1_pro_sharpa_relative_eef_egodex_wrist_only": 26,
29
+ "real_r1_pro_sharpa_relative_eef_maxinsights": 26,
30
+ "real_r1_pro_sharpa_relative_eef_maxinsights_relative_hand": 26,
31
+ "real_r1_pro_sharpa_relative_eef_mecka": 26,
32
+ "real_r1_pro_sharpa_relative_eef_mecka_relative_hand": 26,
33
+ "real_g1_relative_eef_absolute_joints": 25,
34
+ "real_g1_relative_eef_absolute_joints_wrist_cam": 25,
35
+ "real_g1_relative_eef_relative_joints": 25,
36
+ "real_r1_pro_sharpa_relative_eef_relative_hand_relative_joint": 26,
37
+ "real_r1_pro_sharpa_relative_joint": 29,
38
+ "oxe_droid_relative_eef_relative_joint": 24,
39
+ "oxe_droid_relative_eef_relative_joint_swapped": 24,
40
+ "oxe_droid_relative_eef_relative_joint_upweight_z": 24,
41
+ "oxe_droid_relative_eef_relative_joint_upweight_z_swapped": 24,
42
+ "oxe_droid_relative_eef_relative_joint_3view": 24,
43
+ "oxe_droid_relative_eef_relative_joint_3view_swapped": 24,
44
+ "oxe_droid_relative_eef": 24,
45
+ "oxe_droid_joint_position_relative": 24,
46
+ "xdof_relative_eef_relative_joint": 27,
47
+ "xdof_relative_eef_relative_joint_subtask": 27,
48
+ "xdof_relative_eef": 27,
49
+ "xdof_relative_joint": 28,
50
+ "simpler_env_google": 0,
51
+ "simpler_env_widowx": 1,
52
+ "libero_sim": 2,
53
+ "droid_sim": 3,
54
+ "new_embodiment": 10,
55
+ "robocasa_gr1_tabletop": 10,
56
+ "unitree_g1_sonic": 11
57
+ }
SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,248 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ load_config_path: null
2
+ model:
3
+ model_type: Gr00tN1d7
4
+ model_dtype: bfloat16
5
+ model_name: nvidia/Cosmos-Reason2-2B
6
+ backbone_model_type: qwen
7
+ model_revision: null
8
+ tune_top_llm_layers: 0
9
+ backbone_embedding_dim: 2048
10
+ tune_llm: false
11
+ tune_visual: true
12
+ select_layer: 12
13
+ reproject_vision: false
14
+ use_flash_attention: true
15
+ load_bf16: false
16
+ backbone_trainable_params_fp32: true
17
+ image_crop_size:
18
+ - 230
19
+ - 230
20
+ image_target_size:
21
+ - 256
22
+ - 256
23
+ shortest_image_edge: null
24
+ crop_fraction: null
25
+ random_rotation_angle: null
26
+ color_jitter_params:
27
+ brightness: 0.3
28
+ contrast: 0.4
29
+ saturation: 0.5
30
+ hue: 0.08
31
+ use_albumentations_transforms: true
32
+ letter_box_transform: false
33
+ extra_augmentation_config: null
34
+ formalize_language: true
35
+ apply_sincos_state_encoding: false
36
+ use_percentiles: true
37
+ use_relative_action: true
38
+ max_state_dim: 132
39
+ max_action_dim: 132
40
+ action_horizon: 40
41
+ hidden_size: 1024
42
+ input_embedding_dim: 1536
43
+ state_history_length: 1
44
+ add_pos_embed: true
45
+ attn_dropout: 0.2
46
+ use_vlln: true
47
+ max_seq_len: 1024
48
+ use_alternate_vl_dit: true
49
+ attend_text_every_n_blocks: 2
50
+ diffusion_model_cfg:
51
+ positional_embeddings: null
52
+ num_layers: 16
53
+ num_attention_heads: 32
54
+ attention_head_dim: 48
55
+ norm_type: ada_norm
56
+ dropout: 0.2
57
+ final_dropout: true
58
+ output_dim: 1024
59
+ interleave_self_attention: true
60
+ num_inference_timesteps: 4
61
+ noise_beta_alpha: 1.5
62
+ noise_beta_beta: 1.0
63
+ noise_s: 0.999
64
+ num_timestep_buckets: 1000
65
+ tune_projector: true
66
+ tune_diffusion_model: true
67
+ tune_vlln: true
68
+ state_dropout_prob: 0.0
69
+ exclude_state: false
70
+ use_mean_std: false
71
+ max_num_embodiments: 32
72
+ data:
73
+ datasets:
74
+ - dataset_paths:
75
+ - /workspace/datasets/osmo_sft/mimicGen_lerobot
76
+ embodiment_tag: new_embodiment
77
+ mix_ratio: 1.0
78
+ dataset_type: physical_embodiment
79
+ val_dataset_path: null
80
+ modality_configs:
81
+ new_embodiment:
82
+ video:
83
+ delta_indices:
84
+ - 0
85
+ modality_keys:
86
+ - high
87
+ - left_wrist_view
88
+ - right_wrist_view
89
+ sin_cos_embedding_keys: null
90
+ mean_std_embedding_keys: null
91
+ action_configs: null
92
+ state:
93
+ delta_indices:
94
+ - 0
95
+ modality_keys:
96
+ - left_arm
97
+ - right_arm
98
+ - left_hand
99
+ - right_hand
100
+ sin_cos_embedding_keys: null
101
+ mean_std_embedding_keys: null
102
+ action_configs: null
103
+ action:
104
+ delta_indices:
105
+ - 0
106
+ - 1
107
+ - 2
108
+ - 3
109
+ - 4
110
+ - 5
111
+ - 6
112
+ - 7
113
+ - 8
114
+ - 9
115
+ - 10
116
+ - 11
117
+ - 12
118
+ - 13
119
+ - 14
120
+ - 15
121
+ - 16
122
+ - 17
123
+ - 18
124
+ - 19
125
+ - 20
126
+ - 21
127
+ - 22
128
+ - 23
129
+ - 24
130
+ - 25
131
+ - 26
132
+ - 27
133
+ - 28
134
+ - 29
135
+ - 30
136
+ - 31
137
+ modality_keys:
138
+ - left_arm
139
+ - right_arm
140
+ - left_hand
141
+ - right_hand
142
+ sin_cos_embedding_keys: null
143
+ mean_std_embedding_keys: null
144
+ action_configs:
145
+ - rep: ABSOLUTE
146
+ type: NON_EEF
147
+ format: DEFAULT
148
+ state_key: null
149
+ - rep: ABSOLUTE
150
+ type: NON_EEF
151
+ format: DEFAULT
152
+ state_key: null
153
+ - rep: ABSOLUTE
154
+ type: NON_EEF
155
+ format: DEFAULT
156
+ state_key: null
157
+ - rep: ABSOLUTE
158
+ type: NON_EEF
159
+ format: DEFAULT
160
+ state_key: null
161
+ language:
162
+ delta_indices:
163
+ - 0
164
+ modality_keys:
165
+ - annotation.human.task_description
166
+ sin_cos_embedding_keys: null
167
+ mean_std_embedding_keys: null
168
+ action_configs: null
169
+ download_cache: false
170
+ shard_size: 1024
171
+ episode_sampling_rate: 0.1
172
+ num_shards_per_epoch: 100000
173
+ ds_weights_alpha: null
174
+ override_pretraining_statistics: true
175
+ mode: single_turn
176
+ random_chop: 0.0
177
+ mock_dataset_mode: false
178
+ shuffle: true
179
+ seed: 42
180
+ multiprocessing_context: fork
181
+ allow_padding: false
182
+ subsample_ratio: 1.0
183
+ image_crop_size:
184
+ - 244
185
+ - 244
186
+ image_target_size:
187
+ - 224
188
+ - 224
189
+ training:
190
+ output_dir: /osmo/data/output/sft
191
+ experiment_name: pack_agx_mimicgen_sft_fixData-1
192
+ max_steps: 40000
193
+ global_batch_size: 176
194
+ per_gpu_batch_size: null
195
+ gradient_accumulation_steps: 1
196
+ learning_rate: 0.0001
197
+ lr_scheduler_type: cosine
198
+ weight_decay: 1.0e-05
199
+ warmup_ratio: 0.05
200
+ warmup_steps: 0
201
+ max_grad_norm: 1.0
202
+ optim: adamw_torch
203
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
204
+ skip_weight_loading: false
205
+ tf32: true
206
+ fp16: false
207
+ bf16: true
208
+ eval_bf16: true
209
+ logging_steps: 10
210
+ save_steps: 10000
211
+ save_total_limit: 2
212
+ save_vl_model: false
213
+ save_only_model: false
214
+ resume_from_checkpoint: false
215
+ upload_checkpoints: false
216
+ upload_every: 1000
217
+ upload_last_n_checkpoints: 5
218
+ max_concurrent_uploads: 2
219
+ eval_strategy: 'no'
220
+ eval_steps: 500
221
+ eval_set_split_ratio: 0.1
222
+ eval_batch_size: 2
223
+ save_best_eval_metric_name: ''
224
+ save_best_eval_metric_greater_is_better: true
225
+ deepspeed_stage: 2
226
+ gradient_checkpointing: true
227
+ transformers_trust_remote_code: true
228
+ transformers_local_files_only: false
229
+ transformers_cache_dir: null
230
+ transformers_access_token: null
231
+ use_ddp: false
232
+ ddp_bucket_cap_mb: 100
233
+ num_gpus: 8
234
+ dataloader_num_workers: 8
235
+ remove_unused_columns: false
236
+ use_wandb: true
237
+ wandb_project: pack_agx_orin
238
+ enable_profiling: false
239
+ max_retries: 3
240
+ assert_loss_less_than: null
241
+ add_rl_callback: false
242
+ enable_open_loop_eval: false
243
+ open_loop_eval_traj_ids:
244
+ - 0
245
+ open_loop_eval_steps_per_traj: 100
246
+ open_loop_eval_plot_indices: null
247
+ max_steps: 40000
248
+ save_steps: 10000
SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/experiment_cfg/config.yaml ADDED
@@ -0,0 +1,246 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ load_config_path: null
2
+ model:
3
+ model_type: Gr00tN1d7
4
+ model_dtype: bfloat16
5
+ model_name: nvidia/Cosmos-Reason2-2B
6
+ backbone_model_type: qwen
7
+ model_revision: null
8
+ tune_top_llm_layers: 0
9
+ backbone_embedding_dim: 2048
10
+ tune_llm: false
11
+ tune_visual: true
12
+ select_layer: 12
13
+ reproject_vision: false
14
+ use_flash_attention: true
15
+ load_bf16: false
16
+ backbone_trainable_params_fp32: true
17
+ image_crop_size:
18
+ - 230
19
+ - 230
20
+ image_target_size:
21
+ - 256
22
+ - 256
23
+ shortest_image_edge: null
24
+ crop_fraction: null
25
+ random_rotation_angle: null
26
+ color_jitter_params:
27
+ brightness: 0.3
28
+ contrast: 0.4
29
+ saturation: 0.5
30
+ hue: 0.08
31
+ use_albumentations_transforms: true
32
+ letter_box_transform: false
33
+ extra_augmentation_config: null
34
+ formalize_language: true
35
+ apply_sincos_state_encoding: false
36
+ use_percentiles: true
37
+ use_relative_action: true
38
+ max_state_dim: 132
39
+ max_action_dim: 132
40
+ action_horizon: 40
41
+ hidden_size: 1024
42
+ input_embedding_dim: 1536
43
+ state_history_length: 1
44
+ add_pos_embed: true
45
+ attn_dropout: 0.2
46
+ use_vlln: true
47
+ max_seq_len: 1024
48
+ use_alternate_vl_dit: true
49
+ attend_text_every_n_blocks: 2
50
+ diffusion_model_cfg:
51
+ positional_embeddings: null
52
+ num_layers: 16
53
+ num_attention_heads: 32
54
+ attention_head_dim: 48
55
+ norm_type: ada_norm
56
+ dropout: 0.2
57
+ final_dropout: true
58
+ output_dim: 1024
59
+ interleave_self_attention: true
60
+ num_inference_timesteps: 4
61
+ noise_beta_alpha: 1.5
62
+ noise_beta_beta: 1.0
63
+ noise_s: 0.999
64
+ num_timestep_buckets: 1000
65
+ tune_projector: true
66
+ tune_diffusion_model: true
67
+ tune_vlln: true
68
+ state_dropout_prob: 0.0
69
+ exclude_state: false
70
+ use_mean_std: false
71
+ max_num_embodiments: 32
72
+ data:
73
+ datasets:
74
+ - dataset_paths:
75
+ - /workspace/datasets/osmo_sft/mimicGen_lerobot
76
+ embodiment_tag: new_embodiment
77
+ mix_ratio: 1.0
78
+ dataset_type: physical_embodiment
79
+ val_dataset_path: null
80
+ modality_configs:
81
+ new_embodiment:
82
+ video:
83
+ delta_indices:
84
+ - 0
85
+ modality_keys:
86
+ - high
87
+ - left_wrist_view
88
+ - right_wrist_view
89
+ sin_cos_embedding_keys: null
90
+ mean_std_embedding_keys: null
91
+ action_configs: null
92
+ state:
93
+ delta_indices:
94
+ - 0
95
+ modality_keys:
96
+ - left_arm
97
+ - right_arm
98
+ - left_hand
99
+ - right_hand
100
+ sin_cos_embedding_keys: null
101
+ mean_std_embedding_keys: null
102
+ action_configs: null
103
+ action:
104
+ delta_indices:
105
+ - 0
106
+ - 1
107
+ - 2
108
+ - 3
109
+ - 4
110
+ - 5
111
+ - 6
112
+ - 7
113
+ - 8
114
+ - 9
115
+ - 10
116
+ - 11
117
+ - 12
118
+ - 13
119
+ - 14
120
+ - 15
121
+ - 16
122
+ - 17
123
+ - 18
124
+ - 19
125
+ - 20
126
+ - 21
127
+ - 22
128
+ - 23
129
+ - 24
130
+ - 25
131
+ - 26
132
+ - 27
133
+ - 28
134
+ - 29
135
+ - 30
136
+ - 31
137
+ modality_keys:
138
+ - left_arm
139
+ - right_arm
140
+ - left_hand
141
+ - right_hand
142
+ sin_cos_embedding_keys: null
143
+ mean_std_embedding_keys: null
144
+ action_configs:
145
+ - rep: absolute
146
+ type: non_eef
147
+ format: default
148
+ state_key: null
149
+ - rep: absolute
150
+ type: non_eef
151
+ format: default
152
+ state_key: null
153
+ - rep: absolute
154
+ type: non_eef
155
+ format: default
156
+ state_key: null
157
+ - rep: absolute
158
+ type: non_eef
159
+ format: default
160
+ state_key: null
161
+ language:
162
+ delta_indices:
163
+ - 0
164
+ modality_keys:
165
+ - annotation.human.task_description
166
+ sin_cos_embedding_keys: null
167
+ mean_std_embedding_keys: null
168
+ action_configs: null
169
+ download_cache: false
170
+ shard_size: 1024
171
+ episode_sampling_rate: 0.1
172
+ num_shards_per_epoch: 100000
173
+ ds_weights_alpha: null
174
+ override_pretraining_statistics: true
175
+ mode: single_turn
176
+ random_chop: 0.0
177
+ mock_dataset_mode: false
178
+ shuffle: true
179
+ seed: 42
180
+ multiprocessing_context: fork
181
+ allow_padding: false
182
+ subsample_ratio: 1.0
183
+ image_crop_size:
184
+ - 244
185
+ - 244
186
+ image_target_size:
187
+ - 224
188
+ - 224
189
+ training:
190
+ output_dir: /osmo/data/output/sft
191
+ experiment_name: pack_agx_mimicgen_sft_fixData-1
192
+ max_steps: 40000
193
+ global_batch_size: 176
194
+ per_gpu_batch_size: null
195
+ gradient_accumulation_steps: 1
196
+ learning_rate: 0.0001
197
+ lr_scheduler_type: cosine
198
+ weight_decay: 1.0e-05
199
+ warmup_ratio: 0.05
200
+ warmup_steps: 0
201
+ max_grad_norm: 1.0
202
+ optim: adamw_torch
203
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
204
+ skip_weight_loading: false
205
+ tf32: true
206
+ fp16: false
207
+ bf16: true
208
+ eval_bf16: true
209
+ logging_steps: 10
210
+ save_steps: 10000
211
+ save_total_limit: 2
212
+ save_vl_model: false
213
+ save_only_model: false
214
+ resume_from_checkpoint: false
215
+ upload_checkpoints: false
216
+ upload_every: 1000
217
+ upload_last_n_checkpoints: 5
218
+ max_concurrent_uploads: 2
219
+ eval_strategy: 'no'
220
+ eval_steps: 500
221
+ eval_set_split_ratio: 0.1
222
+ eval_batch_size: 2
223
+ save_best_eval_metric_name: ''
224
+ save_best_eval_metric_greater_is_better: true
225
+ deepspeed_stage: 2
226
+ gradient_checkpointing: true
227
+ transformers_trust_remote_code: true
228
+ transformers_local_files_only: false
229
+ transformers_cache_dir: null
230
+ transformers_access_token: null
231
+ use_ddp: false
232
+ ddp_bucket_cap_mb: 100
233
+ num_gpus: 8
234
+ dataloader_num_workers: 8
235
+ remove_unused_columns: false
236
+ use_wandb: true
237
+ wandb_project: pack_agx_orin
238
+ enable_profiling: false
239
+ max_retries: 3
240
+ assert_loss_less_than: null
241
+ add_rl_callback: false
242
+ enable_open_loop_eval: false
243
+ open_loop_eval_traj_ids:
244
+ - 0
245
+ open_loop_eval_steps_per_traj: 100
246
+ open_loop_eval_plot_indices: null
SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,817 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "left_arm": {
5
+ "min": [
6
+ -0.5629809498786926,
7
+ -0.177187979221344,
8
+ -0.22186532616615295,
9
+ -0.505587100982666,
10
+ -0.5124126076698303,
11
+ -0.44763287901878357,
12
+ -0.43886610865592957
13
+ ],
14
+ "max": [
15
+ 0.5412801504135132,
16
+ 0.3043138384819031,
17
+ 0.28761813044548035,
18
+ 1.126184105873108,
19
+ 0.4076734483242035,
20
+ 0.21795514225959778,
21
+ 0.42292702198028564
22
+ ],
23
+ "mean": [
24
+ 0.05245085805654526,
25
+ 0.0588013231754303,
26
+ 0.03832138329744339,
27
+ 0.240767240524292,
28
+ -0.14904314279556274,
29
+ 0.012272846885025501,
30
+ -0.0728105753660202
31
+ ],
32
+ "std": [
33
+ 0.1783587485551834,
34
+ 0.09095463156700134,
35
+ 0.12008842080831528,
36
+ 0.4129796028137207,
37
+ 0.235039621591568,
38
+ 0.11135142296552658,
39
+ 0.18434210121631622
40
+ ],
41
+ "q01": [
42
+ -0.30057936906814575,
43
+ -0.14425581693649292,
44
+ -0.17863908410072327,
45
+ -0.36721596121788025,
46
+ -0.4804217517375946,
47
+ -0.22883346676826477,
48
+ -0.42328712344169617
49
+ ],
50
+ "q99": [
51
+ 0.4457714259624481,
52
+ 0.2627217471599579,
53
+ 0.24426555633544922,
54
+ 0.9861360192298889,
55
+ 0.22885262966156006,
56
+ 0.17035524547100067,
57
+ 0.33738628029823303
58
+ ]
59
+ },
60
+ "right_arm": {
61
+ "min": [
62
+ -0.5351261496543884,
63
+ -0.7679711580276489,
64
+ -0.44770416617393494,
65
+ -0.5488454103469849,
66
+ -1.7459508180618286,
67
+ -0.30524349212646484,
68
+ -0.3152182698249817
69
+ ],
70
+ "max": [
71
+ 0.4151162803173065,
72
+ 0.41699811816215515,
73
+ 1.109545350074768,
74
+ 1.1100053787231445,
75
+ 0.5287642478942871,
76
+ 0.5761488676071167,
77
+ 0.6478164196014404
78
+ ],
79
+ "mean": [
80
+ -0.07371192425489426,
81
+ -0.21649888157844543,
82
+ 0.366228848695755,
83
+ 0.1588103473186493,
84
+ -0.920173704624176,
85
+ 0.3552303910255432,
86
+ 0.1983693689107895
87
+ ],
88
+ "std": [
89
+ 0.1966324895620346,
90
+ 0.23837684094905853,
91
+ 0.3103739321231842,
92
+ 0.2731437087059021,
93
+ 0.5628638863563538,
94
+ 0.17721758782863617,
95
+ 0.2125815898180008
96
+ ],
97
+ "q01": [
98
+ -0.45319539308547974,
99
+ -0.6450024843215942,
100
+ -0.37088584899902344,
101
+ -0.39617177844047546,
102
+ -1.6923842430114746,
103
+ -0.04236690327525139,
104
+ -0.15771058201789856
105
+ ],
106
+ "q99": [
107
+ 0.36341798305511475,
108
+ 0.3209429383277893,
109
+ 0.761161208152771,
110
+ 0.715834379196167,
111
+ 0.44061416387557983,
112
+ 0.5759999752044678,
113
+ 0.6215032339096069
114
+ ]
115
+ },
116
+ "left_hand": {
117
+ "min": [
118
+ 0.3110215663909912,
119
+ -0.17274317145347595,
120
+ -0.3405802845954895,
121
+ -0.26480159163475037,
122
+ 0.048272259533405304,
123
+ -0.061133794486522675,
124
+ -0.19504642486572266,
125
+ 0.04545611888170242,
126
+ 0.11932947486639023,
127
+ -0.07047039270401001,
128
+ -0.07931214570999146,
129
+ 0.07029285281896591,
130
+ 0.08481545001268387,
131
+ -0.07870163023471832,
132
+ -0.04627322033047676,
133
+ 0.09658147394657135,
134
+ 0.1071639284491539,
135
+ 0.1111721470952034,
136
+ -0.10611534118652344,
137
+ -0.054615192115306854,
138
+ 0.11654524505138397,
139
+ 0.06683791428804398
140
+ ],
141
+ "max": [
142
+ 0.749590277671814,
143
+ 0.23624859750270844,
144
+ 0.9072500467300415,
145
+ 0.14956827461719513,
146
+ 0.684272825717926,
147
+ 0.5906409621238708,
148
+ -0.09926573187112808,
149
+ 0.1350926160812378,
150
+ 0.17703042924404144,
151
+ 0.36039674282073975,
152
+ -0.007096846587955952,
153
+ 0.19322505593299866,
154
+ 0.14509114623069763,
155
+ 0.26332560181617737,
156
+ 0.07052827626466751,
157
+ 0.3001532554626465,
158
+ 0.17408396303653717,
159
+ 0.13733766973018646,
160
+ 0.17266049981117249,
161
+ 0.09317488223314285,
162
+ 0.5587858557701111,
163
+ 0.08887965232133865
164
+ ],
165
+ "mean": [
166
+ 0.5027308464050293,
167
+ 0.044371869415044785,
168
+ 0.781243622303009,
169
+ -0.24171887338161469,
170
+ 0.07806016504764557,
171
+ 0.4608381986618042,
172
+ -0.17402991652488708,
173
+ 0.08406005799770355,
174
+ 0.13850891590118408,
175
+ 0.3059825003147125,
176
+ -0.04730428382754326,
177
+ 0.17650458216667175,
178
+ 0.08678989112377167,
179
+ 0.1955358386039734,
180
+ -0.035045262426137924,
181
+ 0.27982884645462036,
182
+ 0.1538132131099701,
183
+ 0.1304696649312973,
184
+ 0.07136118412017822,
185
+ -0.01641261577606201,
186
+ 0.4955330491065979,
187
+ 0.07669968903064728
188
+ ],
189
+ "std": [
190
+ 0.1391039490699768,
191
+ 0.11435980349779129,
192
+ 0.113160640001297,
193
+ 0.03033993579447264,
194
+ 0.0463496595621109,
195
+ 0.06467445194721222,
196
+ 0.0190382432192564,
197
+ 0.020050479099154472,
198
+ 0.02032870426774025,
199
+ 0.04131606966257095,
200
+ 0.008440224453806877,
201
+ 0.012853926047682627,
202
+ 0.0040684076957404345,
203
+ 0.04319994896650314,
204
+ 0.008194883354008198,
205
+ 0.016607539728283778,
206
+ 0.023031415417790413,
207
+ 0.0013840969186277568,
208
+ 0.050044503062963486,
209
+ 0.017094967886805534,
210
+ 0.042078290134668184,
211
+ 0.003900207346305244
212
+ ],
213
+ "q01": [
214
+ 0.3769892454147339,
215
+ -0.16967077553272247,
216
+ 0.45211702585220337,
217
+ -0.26393452286720276,
218
+ 0.04967505484819412,
219
+ 0.30734580755233765,
220
+ -0.19324931502342224,
221
+ 0.06023586913943291,
222
+ 0.1193559542298317,
223
+ 0.2122785598039627,
224
+ -0.07604130357503891,
225
+ 0.1507781744003296,
226
+ 0.08481551706790924,
227
+ 0.13395991921424866,
228
+ -0.03944357484579086,
229
+ 0.24940024316310883,
230
+ 0.1102106124162674,
231
+ 0.12604066729545593,
232
+ 0.015628471970558167,
233
+ -0.037606898695230484,
234
+ 0.3854920566082001,
235
+ 0.06685926765203476
236
+ ],
237
+ "q99": [
238
+ 0.7471694350242615,
239
+ 0.23384490609169006,
240
+ 0.9047397971153259,
241
+ -0.14035801589488983,
242
+ 0.22135315835475922,
243
+ 0.5864326357841492,
244
+ -0.12027500569820404,
245
+ 0.13408620655536652,
246
+ 0.17695677280426025,
247
+ 0.36007148027420044,
248
+ -0.02816118858754635,
249
+ 0.19282959401607513,
250
+ 0.09945125132799149,
251
+ 0.25627684593200684,
252
+ -0.009219160303473473,
253
+ 0.3001519441604614,
254
+ 0.17408394813537598,
255
+ 0.13138802349567413,
256
+ 0.1710493117570877,
257
+ 0.012382262386381626,
258
+ 0.5540529489517212,
259
+ 0.08343399316072464
260
+ ]
261
+ },
262
+ "right_hand": {
263
+ "min": [
264
+ 0.04009956121444702,
265
+ -0.3492163419723511,
266
+ -0.13204336166381836,
267
+ -0.34144726395606995,
268
+ 0.08422064781188965,
269
+ -0.01858464814722538,
270
+ -0.30175480246543884,
271
+ 0.052651457488536835,
272
+ 0.11926719546318054,
273
+ -0.09594082832336426,
274
+ -0.11071047186851501,
275
+ 0.07874276489019394,
276
+ 0.16694247722625732,
277
+ -0.092088982462883,
278
+ -0.10332406312227249,
279
+ 0.05292117968201637,
280
+ 0.12555740773677826,
281
+ -7.358708535321057e-05,
282
+ -0.015641598030924797,
283
+ 0.015677252784371376,
284
+ 0.07094672322273254,
285
+ 0.05056879669427872
286
+ ],
287
+ "max": [
288
+ 0.6687175035476685,
289
+ 0.16571688652038574,
290
+ 0.6953725218772888,
291
+ 0.0049558840692043304,
292
+ 0.42503973841667175,
293
+ 0.643510103225708,
294
+ 0.05673333257436752,
295
+ 0.31882360577583313,
296
+ 0.4645529091358185,
297
+ 0.4012362062931061,
298
+ 0.08867213875055313,
299
+ 0.5379688143730164,
300
+ 0.6116448640823364,
301
+ 0.4138016998767853,
302
+ 0.08890783786773682,
303
+ 0.3182452321052551,
304
+ 0.563942015171051,
305
+ 0.13440823554992676,
306
+ 0.7590939998626709,
307
+ 0.2253386676311493,
308
+ 0.41105303168296814,
309
+ 0.2441820651292801
310
+ ],
311
+ "mean": [
312
+ 0.4645604193210602,
313
+ -0.22611550986766815,
314
+ 0.5470046997070312,
315
+ -0.26729831099510193,
316
+ 0.2005244791507721,
317
+ 0.41302406787872314,
318
+ -0.208993598818779,
319
+ 0.2136479914188385,
320
+ 0.3096477687358856,
321
+ 0.25719523429870605,
322
+ -0.08060095459222794,
323
+ 0.3747420310974121,
324
+ 0.4548353850841522,
325
+ 0.2681923508644104,
326
+ -0.03253396600484848,
327
+ 0.22708161175251007,
328
+ 0.45014604926109314,
329
+ 0.05951449275016785,
330
+ 0.6587040424346924,
331
+ 0.11709203571081161,
332
+ 0.27106180787086487,
333
+ 0.16710738837718964
334
+ ],
335
+ "std": [
336
+ 0.18583320081233978,
337
+ 0.1461419314146042,
338
+ 0.2174922376871109,
339
+ 0.04459349066019058,
340
+ 0.07855303585529327,
341
+ 0.18568646907806396,
342
+ 0.0490199476480484,
343
+ 0.09472297877073288,
344
+ 0.09909675270318985,
345
+ 0.1475902795791626,
346
+ 0.016289083287119865,
347
+ 0.13884447515010834,
348
+ 0.11690541356801987,
349
+ 0.11812587082386017,
350
+ 0.053294721990823746,
351
+ 0.07421745359897614,
352
+ 0.1064753383398056,
353
+ 0.062272630631923676,
354
+ 0.11878073215484619,
355
+ 0.042994096875190735,
356
+ 0.10352463275194168,
357
+ 0.0648803561925888
358
+ ],
359
+ "q01": [
360
+ 0.07583727687597275,
361
+ -0.3491000235080719,
362
+ -0.12692514061927795,
363
+ -0.3308507800102234,
364
+ 0.09154554456472397,
365
+ 0.004721301142126322,
366
+ -0.2881492078304291,
367
+ 0.05265166237950325,
368
+ 0.1209481880068779,
369
+ -0.08444974571466446,
370
+ -0.1107003465294838,
371
+ 0.08550962060689926,
372
+ 0.17066536843776703,
373
+ -0.07877258211374283,
374
+ -0.09157676249742508,
375
+ 0.05840425565838814,
376
+ 0.13074491918087006,
377
+ -1.0690812679570172e-08,
378
+ 0.2881232500076294,
379
+ 0.015767911449074745,
380
+ 0.07359201461076736,
381
+ 0.050768159329891205
382
+ ],
383
+ "q99": [
384
+ 0.664175271987915,
385
+ 0.16195373237133026,
386
+ 0.6796196103096008,
387
+ -0.16884665191173553,
388
+ 0.4249716103076935,
389
+ 0.6342748403549194,
390
+ -0.12000476568937302,
391
+ 0.31848403811454773,
392
+ 0.4630919098854065,
393
+ 0.40073272585868835,
394
+ -0.05453495308756828,
395
+ 0.537603497505188,
396
+ 0.6094125509262085,
397
+ 0.4007859230041504,
398
+ 0.08853022754192352,
399
+ 0.3166349530220032,
400
+ 0.5605353713035583,
401
+ 0.13426418602466583,
402
+ 0.747979998588562,
403
+ 0.2228250950574875,
404
+ 0.40390974283218384,
405
+ 0.24417828023433685
406
+ ]
407
+ }
408
+ },
409
+ "action": {
410
+ "left_arm": {
411
+ "min": [
412
+ -0.3936310410499573,
413
+ -0.18593768775463104,
414
+ -0.23139724135398865,
415
+ -0.5043433904647827,
416
+ -0.5128127932548523,
417
+ -0.359060674905777,
418
+ -0.4394310712814331
419
+ ],
420
+ "max": [
421
+ 0.5497207641601562,
422
+ 0.2946853041648865,
423
+ 0.28778645396232605,
424
+ 1.2456637620925903,
425
+ 0.3696635365486145,
426
+ 0.21236462891101837,
427
+ 0.4390096068382263
428
+ ],
429
+ "mean": [
430
+ 0.05051468685269356,
431
+ 0.05497479438781738,
432
+ 0.03528280183672905,
433
+ 0.23978719115257263,
434
+ -0.1484392285346985,
435
+ 0.008253464475274086,
436
+ -0.07847519218921661
437
+ ],
438
+ "std": [
439
+ 0.17971955239772797,
440
+ 0.09344691038131714,
441
+ 0.12429414689540863,
442
+ 0.41245025396347046,
443
+ 0.23608259856700897,
444
+ 0.11481130868196487,
445
+ 0.1786075085401535
446
+ ],
447
+ "q01": [
448
+ -0.2965664267539978,
449
+ -0.15491817891597748,
450
+ -0.19237348437309265,
451
+ -0.37075141072273254,
452
+ -0.48096907138824463,
453
+ -0.24133530259132385,
454
+ -0.4231419265270233
455
+ ],
456
+ "q99": [
457
+ 0.44708922505378723,
458
+ 0.2624276280403137,
459
+ 0.2452559620141983,
460
+ 0.9885529279708862,
461
+ 0.231220081448555,
462
+ 0.1688399612903595,
463
+ 0.32577186822891235
464
+ ]
465
+ },
466
+ "right_arm": {
467
+ "min": [
468
+ -0.5487637519836426,
469
+ -0.7692788243293762,
470
+ -0.4491795301437378,
471
+ -0.5501343011856079,
472
+ -1.7618770599365234,
473
+ -0.2567974030971527,
474
+ -0.3167533874511719
475
+ ],
476
+ "max": [
477
+ 0.41644051671028137,
478
+ 0.414732962846756,
479
+ 1.192404866218567,
480
+ 1.162872314453125,
481
+ 0.5300719141960144,
482
+ 0.5760312080383301,
483
+ 0.6244572401046753
484
+ ],
485
+ "mean": [
486
+ -0.07847775518894196,
487
+ -0.21507976949214935,
488
+ 0.36580389738082886,
489
+ 0.15532125532627106,
490
+ -0.9226366281509399,
491
+ 0.35693642497062683,
492
+ 0.19445733726024628
493
+ ],
494
+ "std": [
495
+ 0.20359672605991364,
496
+ 0.2399168461561203,
497
+ 0.3121553063392639,
498
+ 0.2797538638114929,
499
+ 0.571170449256897,
500
+ 0.17653939127922058,
501
+ 0.1973046511411667
502
+ ],
503
+ "q01": [
504
+ -0.47475317120552063,
505
+ -0.6472804546356201,
506
+ -0.3789006471633911,
507
+ -0.39843204617500305,
508
+ -1.7095903158187866,
509
+ -0.03737292066216469,
510
+ -0.14629459381103516
511
+ ],
512
+ "q99": [
513
+ 0.364156037569046,
514
+ 0.32124608755111694,
515
+ 0.7604050040245056,
516
+ 0.7429817914962769,
517
+ 0.44749119877815247,
518
+ 0.5759999752044678,
519
+ 0.5955920815467834
520
+ ]
521
+ },
522
+ "left_hand": {
523
+ "min": [
524
+ 0.3918694257736206,
525
+ -0.17544984817504883,
526
+ 0.5742565393447876,
527
+ -0.26477256417274475,
528
+ 0.047149769961833954,
529
+ 0.41272443532943726,
530
+ -0.1980220228433609,
531
+ 0.0594881996512413,
532
+ 0.1185695081949234,
533
+ 0.27439790964126587,
534
+ -0.08423010259866714,
535
+ 0.1644398719072342,
536
+ 0.0840228721499443,
537
+ 0.15344059467315674,
538
+ -0.04130871221423149,
539
+ 0.25055503845214844,
540
+ 0.11074753105640411,
541
+ 0.13054612278938293,
542
+ 0.029649898409843445,
543
+ -0.03919065743684769,
544
+ 0.4713686406612396,
545
+ 0.0660797655582428
546
+ ],
547
+ "max": [
548
+ 0.7500954270362854,
549
+ 0.23579378426074982,
550
+ 0.9043861627578735,
551
+ -0.19977086782455444,
552
+ 0.10726554691791534,
553
+ 0.5934039950370789,
554
+ -0.11681664735078812,
555
+ 0.13553768396377563,
556
+ 0.177893728017807,
557
+ 0.36133188009262085,
558
+ -0.036417875438928604,
559
+ 0.19358991086483002,
560
+ 0.08884952962398529,
561
+ 0.26334381103515625,
562
+ -0.03527446091175079,
563
+ 0.3009660840034485,
564
+ 0.17489220201969147,
565
+ 0.13134565949440002,
566
+ 0.17368391156196594,
567
+ 0.007294907234609127,
568
+ 0.554821789264679,
569
+ 0.08312493562698364
570
+ ],
571
+ "mean": [
572
+ 0.5023679137229919,
573
+ 0.040529921650886536,
574
+ 0.7890689969062805,
575
+ -0.24463574588298798,
576
+ 0.07288816571235657,
577
+ 0.4655190110206604,
578
+ -0.1752176284790039,
579
+ 0.08465147018432617,
580
+ 0.13841411471366882,
581
+ 0.309598833322525,
582
+ -0.04855382442474365,
583
+ 0.17729313671588898,
584
+ 0.08652068674564362,
585
+ 0.2009265422821045,
586
+ -0.03867093846201897,
587
+ 0.2823554277420044,
588
+ 0.15447111427783966,
589
+ 0.13080471754074097,
590
+ 0.0728967934846878,
591
+ -0.018348844721913338,
592
+ 0.49855995178222656,
593
+ 0.07631561160087585
594
+ ],
595
+ "std": [
596
+ 0.13876406848430634,
597
+ 0.11674734205007553,
598
+ 0.08069681376218761,
599
+ 0.014197190292179553,
600
+ 0.021397296339273453,
601
+ 0.05358625948429108,
602
+ 0.019477829337120056,
603
+ 0.019863715395331383,
604
+ 0.02082676626741886,
605
+ 0.032544344663619995,
606
+ 0.008071187883615494,
607
+ 0.011100506410002552,
608
+ 0.0013812592951580832,
609
+ 0.04264961555600166,
610
+ 0.001251676701940634,
611
+ 0.011455746367573587,
612
+ 0.022764969617128372,
613
+ 0.00011395005276853349,
614
+ 0.04856700822710991,
615
+ 0.015242788940668106,
616
+ 0.03271668031811693,
617
+ 0.003983301110565662
618
+ ],
619
+ "q01": [
620
+ 0.3929607570171356,
621
+ -0.17535732686519623,
622
+ 0.5742722749710083,
623
+ -0.2647572159767151,
624
+ 0.047164663672447205,
625
+ 0.4127298593521118,
626
+ -0.19425080716609955,
627
+ 0.059504516422748566,
628
+ 0.11858192831277847,
629
+ 0.274588018655777,
630
+ -0.07925879955291748,
631
+ 0.16444341838359833,
632
+ 0.0840383917093277,
633
+ 0.15349750220775604,
634
+ -0.04087769240140915,
635
+ 0.250562846660614,
636
+ 0.11118252575397491,
637
+ 0.1305932104587555,
638
+ 0.030105693265795708,
639
+ -0.039142828434705734,
640
+ 0.4713757038116455,
641
+ 0.06610637158155441
642
+ ],
643
+ "q99": [
644
+ 0.746482253074646,
645
+ 0.23534412682056427,
646
+ 0.9043689370155334,
647
+ -0.19979244470596313,
648
+ 0.10725811123847961,
649
+ 0.5912734270095825,
650
+ -0.11782712489366531,
651
+ 0.13544833660125732,
652
+ 0.17776291072368622,
653
+ 0.3613196015357971,
654
+ -0.03777790069580078,
655
+ 0.19358032941818237,
656
+ 0.08880164474248886,
657
+ 0.26332810521125793,
658
+ -0.0365343913435936,
659
+ 0.30095863342285156,
660
+ 0.17485184967517853,
661
+ 0.1311883181333542,
662
+ 0.17366233468055725,
663
+ 0.0039198752492666245,
664
+ 0.5548086166381836,
665
+ 0.08310144394636154
666
+ ]
667
+ },
668
+ "right_hand": {
669
+ "min": [
670
+ 0.05656527355313301,
671
+ -0.31667423248291016,
672
+ -0.14167125523090363,
673
+ -0.3035319447517395,
674
+ 0.10522574931383133,
675
+ -0.03028453327715397,
676
+ -0.28060656785964966,
677
+ 0.051752351224422455,
678
+ 0.11661845445632935,
679
+ -0.10934512317180634,
680
+ -0.11242455244064331,
681
+ 0.06948084384202957,
682
+ 0.158926323056221,
683
+ -0.10716329514980316,
684
+ -0.08241944760084152,
685
+ 0.048943858593702316,
686
+ 0.1167350709438324,
687
+ 0.02247517742216587,
688
+ 0.27274423837661743,
689
+ 0.014368311502039433,
690
+ 0.06936880946159363,
691
+ 0.04981590062379837
692
+ ],
693
+ "max": [
694
+ 0.7023979425430298,
695
+ 0.16711464524269104,
696
+ 0.7318638563156128,
697
+ -0.20320987701416016,
698
+ 0.5249636173248291,
699
+ 0.643631100654602,
700
+ -0.151748389005661,
701
+ 0.31930866837501526,
702
+ 0.4663243591785431,
703
+ 0.40068960189819336,
704
+ -0.05086810886859894,
705
+ 0.5383633375167847,
706
+ 0.6140285134315491,
707
+ 0.4134733974933624,
708
+ 0.09306606650352478,
709
+ 0.32129883766174316,
710
+ 0.5657610297203064,
711
+ 0.1346166580915451,
712
+ 0.7374468445777893,
713
+ 0.22712412476539612,
714
+ 0.4027511477470398,
715
+ 0.24494193494319916
716
+ ],
717
+ "mean": [
718
+ 0.4848066568374634,
719
+ -0.19299572706222534,
720
+ 0.5662363171577454,
721
+ -0.2588456869125366,
722
+ 0.20845799148082733,
723
+ 0.42852792143821716,
724
+ -0.2047809660434723,
725
+ 0.21921564638614655,
726
+ 0.3092115521430969,
727
+ 0.2600976526737213,
728
+ -0.08231294900178909,
729
+ 0.3772304356098175,
730
+ 0.4559083878993988,
731
+ 0.28397077322006226,
732
+ -0.0328189916908741,
733
+ 0.23260822892189026,
734
+ 0.45114240050315857,
735
+ 0.0716049000620842,
736
+ 0.6632872223854065,
737
+ 0.08630851656198502,
738
+ 0.27243128418922424,
739
+ 0.16675002872943878
740
+ ],
741
+ "std": [
742
+ 0.2009204477071762,
743
+ 0.1260128766298294,
744
+ 0.23686008155345917,
745
+ 0.029346615076065063,
746
+ 0.08104271441698074,
747
+ 0.19370564818382263,
748
+ 0.04986895993351936,
749
+ 0.09879165887832642,
750
+ 0.10072077065706253,
751
+ 0.14988161623477936,
752
+ 0.013688141480088234,
753
+ 0.13982421159744263,
754
+ 0.117080457508564,
755
+ 0.12859608232975006,
756
+ 0.05258253216743469,
757
+ 0.07852780818939209,
758
+ 0.10716189444065094,
759
+ 0.052044015377759933,
760
+ 0.10737068951129913,
761
+ 0.05219526216387749,
762
+ 0.1040802150964737,
763
+ 0.06587769836187363
764
+ ],
765
+ "q01": [
766
+ 0.06780627369880676,
767
+ -0.3166660964488983,
768
+ -0.13796158134937286,
769
+ -0.30219724774360657,
770
+ 0.10594586282968521,
771
+ -0.013602913357317448,
772
+ -0.2792985439300537,
773
+ 0.05178355425596237,
774
+ 0.11790037900209427,
775
+ -0.10453744977712631,
776
+ -0.1124066486954689,
777
+ 0.07422985881567001,
778
+ 0.16193227469921112,
779
+ -0.09863691031932831,
780
+ -0.0823676809668541,
781
+ 0.051408782601356506,
782
+ 0.12347953021526337,
783
+ 0.02248801849782467,
784
+ 0.27782076597213745,
785
+ 0.014379582367837429,
786
+ 0.06937909126281738,
787
+ 0.04985225573182106
788
+ ],
789
+ "q99": [
790
+ 0.702386736869812,
791
+ 0.16283226013183594,
792
+ 0.7304694652557373,
793
+ -0.20540979504585266,
794
+ 0.42579561471939087,
795
+ 0.6435886025428772,
796
+ -0.1518232524394989,
797
+ 0.3192494511604309,
798
+ 0.4635205864906311,
799
+ 0.4006670415401459,
800
+ -0.053987760096788406,
801
+ 0.538344144821167,
802
+ 0.6100354790687561,
803
+ 0.4134240746498108,
804
+ 0.08971802145242691,
805
+ 0.3174830377101898,
806
+ 0.5614163875579834,
807
+ 0.13442933559417725,
808
+ 0.7374396920204163,
809
+ 0.2251783162355423,
810
+ 0.4027312994003296,
811
+ 0.24489758908748627
812
+ ]
813
+ }
814
+ },
815
+ "relative_action": {}
816
+ }
817
+ }
SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/experiment_cfg/final_model_config.json ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "Gr00tN1d7",
3
+ "model_dtype": "bfloat16",
4
+ "model_name": "nvidia/Cosmos-Reason2-2B",
5
+ "backbone_model_type": "qwen",
6
+ "model_revision": null,
7
+ "tune_top_llm_layers": 0,
8
+ "backbone_embedding_dim": 2048,
9
+ "tune_llm": false,
10
+ "tune_visual": true,
11
+ "select_layer": 16,
12
+ "reproject_vision": false,
13
+ "use_flash_attention": true,
14
+ "load_bf16": false,
15
+ "backbone_trainable_params_fp32": true,
16
+ "letter_box_transform": false,
17
+ "extra_augmentation_config": null,
18
+ "apply_sincos_state_encoding": false,
19
+ "use_percentiles": true,
20
+ "use_relative_action": false,
21
+ "max_state_dim": 132,
22
+ "max_action_dim": 132,
23
+ "action_horizon": 40,
24
+ "hidden_size": 1024,
25
+ "input_embedding_dim": 1536,
26
+ "state_history_length": 1,
27
+ "add_pos_embed": true,
28
+ "attn_dropout": 0.2,
29
+ "use_vlln": true,
30
+ "max_seq_len": 1024,
31
+ "use_alternate_vl_dit": true,
32
+ "attend_text_every_n_blocks": 2,
33
+ "diffusion_model_cfg": {
34
+ "attention_head_dim": 48,
35
+ "dropout": 0.2,
36
+ "final_dropout": true,
37
+ "interleave_self_attention": true,
38
+ "norm_type": "ada_norm",
39
+ "num_attention_heads": 32,
40
+ "num_layers": 32,
41
+ "output_dim": 1024,
42
+ "positional_embeddings": null
43
+ },
44
+ "num_inference_timesteps": 4,
45
+ "noise_beta_alpha": 1.5,
46
+ "noise_beta_beta": 1.0,
47
+ "noise_s": 0.999,
48
+ "num_timestep_buckets": 1000,
49
+ "tune_projector": true,
50
+ "tune_diffusion_model": true,
51
+ "tune_vlln": true,
52
+ "state_dropout_prob": 0.0,
53
+ "exclude_state": false,
54
+ "use_mean_std": false,
55
+ "max_num_embodiments": 32
56
+ }
SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/experiment_cfg/final_processor_config.json ADDED
The diff for this file is too large to render. See raw diff
 
SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4a0936f8858ee607f09bf2b35d4cf8d0f3bd4042355e898bf21138871266804
3
+ size 4990519232
SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d047f4c58b6023f487eaa06b962f299ceca2b46e8892ce42237f81af159f5c1a
3
+ size 1919980184
SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/processor_config.json ADDED
@@ -0,0 +1,1156 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "processor_class": "Gr00tN1d7Processor",
3
+ "processor_kwargs": {
4
+ "modality_configs": {
5
+ "real_g1_relative_eef_relative_joints": {
6
+ "video": {
7
+ "delta_indices": [
8
+ -20,
9
+ 0
10
+ ],
11
+ "modality_keys": [
12
+ "ego_view"
13
+ ],
14
+ "sin_cos_embedding_keys": null,
15
+ "mean_std_embedding_keys": null,
16
+ "action_configs": null
17
+ },
18
+ "state": {
19
+ "delta_indices": [
20
+ 0
21
+ ],
22
+ "modality_keys": [
23
+ "left_wrist_eef_9d",
24
+ "right_wrist_eef_9d",
25
+ "left_hand",
26
+ "right_hand",
27
+ "left_arm",
28
+ "right_arm",
29
+ "waist"
30
+ ],
31
+ "sin_cos_embedding_keys": null,
32
+ "mean_std_embedding_keys": null,
33
+ "action_configs": null
34
+ },
35
+ "action": {
36
+ "delta_indices": [
37
+ 0,
38
+ 1,
39
+ 2,
40
+ 3,
41
+ 4,
42
+ 5,
43
+ 6,
44
+ 7,
45
+ 8,
46
+ 9,
47
+ 10,
48
+ 11,
49
+ 12,
50
+ 13,
51
+ 14,
52
+ 15,
53
+ 16,
54
+ 17,
55
+ 18,
56
+ 19,
57
+ 20,
58
+ 21,
59
+ 22,
60
+ 23,
61
+ 24,
62
+ 25,
63
+ 26,
64
+ 27,
65
+ 28,
66
+ 29,
67
+ 30,
68
+ 31,
69
+ 32,
70
+ 33,
71
+ 34,
72
+ 35,
73
+ 36,
74
+ 37,
75
+ 38,
76
+ 39
77
+ ],
78
+ "modality_keys": [
79
+ "left_wrist_eef_9d",
80
+ "right_wrist_eef_9d",
81
+ "left_hand",
82
+ "right_hand",
83
+ "left_arm",
84
+ "right_arm",
85
+ "waist",
86
+ "base_height_command",
87
+ "navigate_command"
88
+ ],
89
+ "sin_cos_embedding_keys": null,
90
+ "mean_std_embedding_keys": null,
91
+ "action_configs": [
92
+ {
93
+ "rep": "RELATIVE",
94
+ "type": "EEF",
95
+ "format": "XYZ_ROT6D",
96
+ "state_key": "left_wrist_eef_9d"
97
+ },
98
+ {
99
+ "rep": "RELATIVE",
100
+ "type": "EEF",
101
+ "format": "XYZ_ROT6D",
102
+ "state_key": "right_wrist_eef_9d"
103
+ },
104
+ {
105
+ "rep": "ABSOLUTE",
106
+ "type": "NON_EEF",
107
+ "format": "DEFAULT",
108
+ "state_key": "left_hand"
109
+ },
110
+ {
111
+ "rep": "ABSOLUTE",
112
+ "type": "NON_EEF",
113
+ "format": "DEFAULT",
114
+ "state_key": "right_hand"
115
+ },
116
+ {
117
+ "rep": "RELATIVE",
118
+ "type": "NON_EEF",
119
+ "format": "DEFAULT",
120
+ "state_key": "left_arm"
121
+ },
122
+ {
123
+ "rep": "RELATIVE",
124
+ "type": "NON_EEF",
125
+ "format": "DEFAULT",
126
+ "state_key": "right_arm"
127
+ },
128
+ {
129
+ "rep": "ABSOLUTE",
130
+ "type": "NON_EEF",
131
+ "format": "DEFAULT",
132
+ "state_key": "waist"
133
+ },
134
+ {
135
+ "rep": "ABSOLUTE",
136
+ "type": "NON_EEF",
137
+ "format": "DEFAULT",
138
+ "state_key": "base_height_command"
139
+ },
140
+ {
141
+ "rep": "ABSOLUTE",
142
+ "type": "NON_EEF",
143
+ "format": "DEFAULT",
144
+ "state_key": "navigate_command"
145
+ }
146
+ ]
147
+ },
148
+ "language": {
149
+ "delta_indices": [
150
+ 0
151
+ ],
152
+ "modality_keys": [
153
+ "annotation.human.task_description"
154
+ ],
155
+ "sin_cos_embedding_keys": null,
156
+ "mean_std_embedding_keys": null,
157
+ "action_configs": null
158
+ }
159
+ },
160
+ "real_r1_pro_sharpa_relative_eef_mecka": {
161
+ "video": {
162
+ "delta_indices": [
163
+ -30,
164
+ 0
165
+ ],
166
+ "modality_keys": [
167
+ "ego_view_cropratio_res320x240_freq30"
168
+ ],
169
+ "sin_cos_embedding_keys": null,
170
+ "mean_std_embedding_keys": null,
171
+ "action_configs": null
172
+ },
173
+ "state": {
174
+ "delta_indices": [
175
+ 0
176
+ ],
177
+ "modality_keys": [
178
+ "left_wrist_eef",
179
+ "right_wrist_eef",
180
+ "left_hand_joints",
181
+ "right_hand_joints"
182
+ ],
183
+ "sin_cos_embedding_keys": null,
184
+ "mean_std_embedding_keys": null,
185
+ "action_configs": null
186
+ },
187
+ "action": {
188
+ "delta_indices": [
189
+ 0,
190
+ 1,
191
+ 2,
192
+ 3,
193
+ 4,
194
+ 5,
195
+ 6,
196
+ 7,
197
+ 8,
198
+ 9,
199
+ 10,
200
+ 11,
201
+ 12,
202
+ 13,
203
+ 14,
204
+ 15,
205
+ 16,
206
+ 17,
207
+ 18,
208
+ 19,
209
+ 20,
210
+ 21,
211
+ 22,
212
+ 23,
213
+ 24,
214
+ 25,
215
+ 26,
216
+ 27,
217
+ 28,
218
+ 29,
219
+ 30,
220
+ 31,
221
+ 32,
222
+ 33,
223
+ 34,
224
+ 35,
225
+ 36,
226
+ 37,
227
+ 38,
228
+ 39
229
+ ],
230
+ "modality_keys": [
231
+ "left_wrist_eef",
232
+ "right_wrist_eef",
233
+ "left_hand_joints",
234
+ "right_hand_joints"
235
+ ],
236
+ "sin_cos_embedding_keys": null,
237
+ "mean_std_embedding_keys": null,
238
+ "action_configs": [
239
+ {
240
+ "rep": "RELATIVE",
241
+ "type": "EEF",
242
+ "format": "XYZ_ROT6D",
243
+ "state_key": "left_wrist_eef"
244
+ },
245
+ {
246
+ "rep": "RELATIVE",
247
+ "type": "EEF",
248
+ "format": "XYZ_ROT6D",
249
+ "state_key": "right_wrist_eef"
250
+ },
251
+ {
252
+ "rep": "ABSOLUTE",
253
+ "type": "NON_EEF",
254
+ "format": "DEFAULT",
255
+ "state_key": "left_hand_joints"
256
+ },
257
+ {
258
+ "rep": "ABSOLUTE",
259
+ "type": "NON_EEF",
260
+ "format": "DEFAULT",
261
+ "state_key": "right_hand_joints"
262
+ }
263
+ ]
264
+ },
265
+ "language": {
266
+ "delta_indices": [
267
+ 0
268
+ ],
269
+ "modality_keys": [
270
+ "annotation.human.coarse_action"
271
+ ],
272
+ "sin_cos_embedding_keys": null,
273
+ "mean_std_embedding_keys": null,
274
+ "action_configs": null
275
+ }
276
+ },
277
+ "real_r1_pro_sharpa_relative_eef_human": {
278
+ "video": {
279
+ "delta_indices": [
280
+ -20,
281
+ 0
282
+ ],
283
+ "modality_keys": [
284
+ "ego_view_res320x240_freq20",
285
+ "left_wrist_view_res320x240_freq20",
286
+ "right_wrist_view_res320x240_freq20"
287
+ ],
288
+ "sin_cos_embedding_keys": null,
289
+ "mean_std_embedding_keys": null,
290
+ "action_configs": null
291
+ },
292
+ "state": {
293
+ "delta_indices": [
294
+ 0
295
+ ],
296
+ "modality_keys": [
297
+ "left_wrist_eef",
298
+ "right_wrist_eef",
299
+ "left_hand_joints",
300
+ "right_hand_joints"
301
+ ],
302
+ "sin_cos_embedding_keys": null,
303
+ "mean_std_embedding_keys": null,
304
+ "action_configs": null
305
+ },
306
+ "action": {
307
+ "delta_indices": [
308
+ 0,
309
+ 1,
310
+ 2,
311
+ 3,
312
+ 4,
313
+ 5,
314
+ 6,
315
+ 7,
316
+ 8,
317
+ 9,
318
+ 10,
319
+ 11,
320
+ 12,
321
+ 13,
322
+ 14,
323
+ 15,
324
+ 16,
325
+ 17,
326
+ 18,
327
+ 19,
328
+ 20,
329
+ 21,
330
+ 22,
331
+ 23,
332
+ 24,
333
+ 25,
334
+ 26,
335
+ 27,
336
+ 28,
337
+ 29,
338
+ 30,
339
+ 31,
340
+ 32,
341
+ 33,
342
+ 34,
343
+ 35,
344
+ 36,
345
+ 37,
346
+ 38,
347
+ 39
348
+ ],
349
+ "modality_keys": [
350
+ "left_wrist_eef",
351
+ "right_wrist_eef",
352
+ "left_hand_joints",
353
+ "right_hand_joints"
354
+ ],
355
+ "sin_cos_embedding_keys": null,
356
+ "mean_std_embedding_keys": null,
357
+ "action_configs": [
358
+ {
359
+ "rep": "RELATIVE",
360
+ "type": "EEF",
361
+ "format": "XYZ_ROT6D",
362
+ "state_key": "left_wrist_eef"
363
+ },
364
+ {
365
+ "rep": "RELATIVE",
366
+ "type": "EEF",
367
+ "format": "XYZ_ROT6D",
368
+ "state_key": "right_wrist_eef"
369
+ },
370
+ {
371
+ "rep": "ABSOLUTE",
372
+ "type": "NON_EEF",
373
+ "format": "DEFAULT",
374
+ "state_key": "left_hand_joints"
375
+ },
376
+ {
377
+ "rep": "ABSOLUTE",
378
+ "type": "NON_EEF",
379
+ "format": "DEFAULT",
380
+ "state_key": "right_hand_joints"
381
+ }
382
+ ]
383
+ },
384
+ "language": {
385
+ "delta_indices": [
386
+ 0
387
+ ],
388
+ "modality_keys": [
389
+ "annotation.human.coarse_action"
390
+ ],
391
+ "sin_cos_embedding_keys": null,
392
+ "mean_std_embedding_keys": null,
393
+ "action_configs": null
394
+ }
395
+ },
396
+ "real_r1_pro_sharpa_relative_eef": {
397
+ "video": {
398
+ "delta_indices": [
399
+ -20,
400
+ 0
401
+ ],
402
+ "modality_keys": [
403
+ "ego_view_res320x240_freq20",
404
+ "left_wrist_view_res320x240_freq20",
405
+ "right_wrist_view_res320x240_freq20"
406
+ ],
407
+ "sin_cos_embedding_keys": null,
408
+ "mean_std_embedding_keys": null,
409
+ "action_configs": null
410
+ },
411
+ "state": {
412
+ "delta_indices": [
413
+ 0
414
+ ],
415
+ "modality_keys": [
416
+ "left_wrist_eef",
417
+ "right_wrist_eef",
418
+ "left_hand_joints",
419
+ "right_hand_joints"
420
+ ],
421
+ "sin_cos_embedding_keys": null,
422
+ "mean_std_embedding_keys": null,
423
+ "action_configs": null
424
+ },
425
+ "action": {
426
+ "delta_indices": [
427
+ 0,
428
+ 1,
429
+ 2,
430
+ 3,
431
+ 4,
432
+ 5,
433
+ 6,
434
+ 7,
435
+ 8,
436
+ 9,
437
+ 10,
438
+ 11,
439
+ 12,
440
+ 13,
441
+ 14,
442
+ 15,
443
+ 16,
444
+ 17,
445
+ 18,
446
+ 19,
447
+ 20,
448
+ 21,
449
+ 22,
450
+ 23,
451
+ 24,
452
+ 25,
453
+ 26,
454
+ 27,
455
+ 28,
456
+ 29,
457
+ 30,
458
+ 31,
459
+ 32,
460
+ 33,
461
+ 34,
462
+ 35,
463
+ 36,
464
+ 37,
465
+ 38,
466
+ 39
467
+ ],
468
+ "modality_keys": [
469
+ "left_wrist_eef",
470
+ "right_wrist_eef",
471
+ "left_hand_joints",
472
+ "right_hand_joints"
473
+ ],
474
+ "sin_cos_embedding_keys": null,
475
+ "mean_std_embedding_keys": null,
476
+ "action_configs": [
477
+ {
478
+ "rep": "RELATIVE",
479
+ "type": "EEF",
480
+ "format": "XYZ_ROT6D",
481
+ "state_key": "left_wrist_eef"
482
+ },
483
+ {
484
+ "rep": "RELATIVE",
485
+ "type": "EEF",
486
+ "format": "XYZ_ROT6D",
487
+ "state_key": "right_wrist_eef"
488
+ },
489
+ {
490
+ "rep": "ABSOLUTE",
491
+ "type": "NON_EEF",
492
+ "format": "DEFAULT",
493
+ "state_key": "left_hand_joints"
494
+ },
495
+ {
496
+ "rep": "ABSOLUTE",
497
+ "type": "NON_EEF",
498
+ "format": "DEFAULT",
499
+ "state_key": "right_hand_joints"
500
+ }
501
+ ]
502
+ },
503
+ "language": {
504
+ "delta_indices": [
505
+ 0
506
+ ],
507
+ "modality_keys": [
508
+ "annotation.human.coarse_action"
509
+ ],
510
+ "sin_cos_embedding_keys": null,
511
+ "mean_std_embedding_keys": null,
512
+ "action_configs": null
513
+ }
514
+ },
515
+ "xdof_relative_eef_relative_joint": {
516
+ "video": {
517
+ "delta_indices": [
518
+ -30,
519
+ 0
520
+ ],
521
+ "modality_keys": [
522
+ "top_camera-images-rgb_320_240",
523
+ "left_camera-images-rgb_320_240",
524
+ "right_camera-images-rgb_320_240"
525
+ ],
526
+ "sin_cos_embedding_keys": null,
527
+ "mean_std_embedding_keys": null,
528
+ "action_configs": null
529
+ },
530
+ "state": {
531
+ "delta_indices": [
532
+ 0
533
+ ],
534
+ "modality_keys": [
535
+ "left_wrist_eef",
536
+ "right_wrist_eef",
537
+ "left_gripper_pos",
538
+ "right_gripper_pos",
539
+ "left_joint_pos",
540
+ "right_joint_pos"
541
+ ],
542
+ "sin_cos_embedding_keys": null,
543
+ "mean_std_embedding_keys": null,
544
+ "action_configs": null
545
+ },
546
+ "action": {
547
+ "delta_indices": [
548
+ 0,
549
+ 1,
550
+ 2,
551
+ 3,
552
+ 4,
553
+ 5,
554
+ 6,
555
+ 7,
556
+ 8,
557
+ 9,
558
+ 10,
559
+ 11,
560
+ 12,
561
+ 13,
562
+ 14,
563
+ 15,
564
+ 16,
565
+ 17,
566
+ 18,
567
+ 19,
568
+ 20,
569
+ 21,
570
+ 22,
571
+ 23,
572
+ 24,
573
+ 25,
574
+ 26,
575
+ 27,
576
+ 28,
577
+ 29,
578
+ 30,
579
+ 31,
580
+ 32,
581
+ 33,
582
+ 34,
583
+ 35,
584
+ 36,
585
+ 37,
586
+ 38,
587
+ 39
588
+ ],
589
+ "modality_keys": [
590
+ "left_wrist_eef",
591
+ "right_wrist_eef",
592
+ "left_gripper_pos",
593
+ "right_gripper_pos",
594
+ "left_joint_pos",
595
+ "right_joint_pos"
596
+ ],
597
+ "sin_cos_embedding_keys": null,
598
+ "mean_std_embedding_keys": null,
599
+ "action_configs": [
600
+ {
601
+ "rep": "RELATIVE",
602
+ "type": "EEF",
603
+ "format": "XYZ_ROT6D",
604
+ "state_key": "left_wrist_eef"
605
+ },
606
+ {
607
+ "rep": "RELATIVE",
608
+ "type": "EEF",
609
+ "format": "XYZ_ROT6D",
610
+ "state_key": "right_wrist_eef"
611
+ },
612
+ {
613
+ "rep": "ABSOLUTE",
614
+ "type": "NON_EEF",
615
+ "format": "DEFAULT",
616
+ "state_key": "left_gripper_pos"
617
+ },
618
+ {
619
+ "rep": "ABSOLUTE",
620
+ "type": "NON_EEF",
621
+ "format": "DEFAULT",
622
+ "state_key": "right_gripper_pos"
623
+ },
624
+ {
625
+ "rep": "RELATIVE",
626
+ "type": "NON_EEF",
627
+ "format": "DEFAULT",
628
+ "state_key": "left_joint_pos"
629
+ },
630
+ {
631
+ "rep": "RELATIVE",
632
+ "type": "NON_EEF",
633
+ "format": "DEFAULT",
634
+ "state_key": "right_joint_pos"
635
+ }
636
+ ]
637
+ },
638
+ "language": {
639
+ "delta_indices": [
640
+ 0
641
+ ],
642
+ "modality_keys": [
643
+ "annotation.task"
644
+ ],
645
+ "sin_cos_embedding_keys": null,
646
+ "mean_std_embedding_keys": null,
647
+ "action_configs": null
648
+ }
649
+ },
650
+ "real_r1_pro_sharpa_relative_eef_maxinsights": {
651
+ "video": {
652
+ "delta_indices": [
653
+ -30,
654
+ 0
655
+ ],
656
+ "modality_keys": [
657
+ "ego_view_cropratio_res320x240_freq30"
658
+ ],
659
+ "sin_cos_embedding_keys": null,
660
+ "mean_std_embedding_keys": null,
661
+ "action_configs": null
662
+ },
663
+ "state": {
664
+ "delta_indices": [
665
+ 0
666
+ ],
667
+ "modality_keys": [
668
+ "left_wrist_eef",
669
+ "right_wrist_eef",
670
+ "left_hand_joints",
671
+ "right_hand_joints"
672
+ ],
673
+ "sin_cos_embedding_keys": null,
674
+ "mean_std_embedding_keys": null,
675
+ "action_configs": null
676
+ },
677
+ "action": {
678
+ "delta_indices": [
679
+ 0,
680
+ 1,
681
+ 2,
682
+ 3,
683
+ 4,
684
+ 5,
685
+ 6,
686
+ 7,
687
+ 8,
688
+ 9,
689
+ 10,
690
+ 11,
691
+ 12,
692
+ 13,
693
+ 14,
694
+ 15,
695
+ 16,
696
+ 17,
697
+ 18,
698
+ 19,
699
+ 20,
700
+ 21,
701
+ 22,
702
+ 23,
703
+ 24,
704
+ 25,
705
+ 26,
706
+ 27,
707
+ 28,
708
+ 29,
709
+ 30,
710
+ 31,
711
+ 32,
712
+ 33,
713
+ 34,
714
+ 35,
715
+ 36,
716
+ 37,
717
+ 38,
718
+ 39
719
+ ],
720
+ "modality_keys": [
721
+ "left_wrist_eef",
722
+ "right_wrist_eef",
723
+ "left_hand_joints",
724
+ "right_hand_joints"
725
+ ],
726
+ "sin_cos_embedding_keys": null,
727
+ "mean_std_embedding_keys": null,
728
+ "action_configs": [
729
+ {
730
+ "rep": "RELATIVE",
731
+ "type": "EEF",
732
+ "format": "XYZ_ROT6D",
733
+ "state_key": "left_wrist_eef"
734
+ },
735
+ {
736
+ "rep": "RELATIVE",
737
+ "type": "EEF",
738
+ "format": "XYZ_ROT6D",
739
+ "state_key": "right_wrist_eef"
740
+ },
741
+ {
742
+ "rep": "ABSOLUTE",
743
+ "type": "NON_EEF",
744
+ "format": "DEFAULT",
745
+ "state_key": "left_hand_joints"
746
+ },
747
+ {
748
+ "rep": "ABSOLUTE",
749
+ "type": "NON_EEF",
750
+ "format": "DEFAULT",
751
+ "state_key": "right_hand_joints"
752
+ }
753
+ ]
754
+ },
755
+ "language": {
756
+ "delta_indices": [
757
+ 0
758
+ ],
759
+ "modality_keys": [
760
+ "annotation.human.coarse_action"
761
+ ],
762
+ "sin_cos_embedding_keys": null,
763
+ "mean_std_embedding_keys": null,
764
+ "action_configs": null
765
+ }
766
+ },
767
+ "xdof_relative_eef_relative_joint_subtask": {
768
+ "video": {
769
+ "delta_indices": [
770
+ -30,
771
+ 0
772
+ ],
773
+ "modality_keys": [
774
+ "top_camera-images-rgb_320_240",
775
+ "left_camera-images-rgb_320_240",
776
+ "right_camera-images-rgb_320_240"
777
+ ],
778
+ "sin_cos_embedding_keys": null,
779
+ "mean_std_embedding_keys": null,
780
+ "action_configs": null
781
+ },
782
+ "state": {
783
+ "delta_indices": [
784
+ 0
785
+ ],
786
+ "modality_keys": [
787
+ "left_wrist_eef",
788
+ "right_wrist_eef",
789
+ "left_gripper_pos",
790
+ "right_gripper_pos",
791
+ "left_joint_pos",
792
+ "right_joint_pos"
793
+ ],
794
+ "sin_cos_embedding_keys": null,
795
+ "mean_std_embedding_keys": null,
796
+ "action_configs": null
797
+ },
798
+ "action": {
799
+ "delta_indices": [
800
+ 0,
801
+ 1,
802
+ 2,
803
+ 3,
804
+ 4,
805
+ 5,
806
+ 6,
807
+ 7,
808
+ 8,
809
+ 9,
810
+ 10,
811
+ 11,
812
+ 12,
813
+ 13,
814
+ 14,
815
+ 15,
816
+ 16,
817
+ 17,
818
+ 18,
819
+ 19,
820
+ 20,
821
+ 21,
822
+ 22,
823
+ 23,
824
+ 24,
825
+ 25,
826
+ 26,
827
+ 27,
828
+ 28,
829
+ 29,
830
+ 30,
831
+ 31,
832
+ 32,
833
+ 33,
834
+ 34,
835
+ 35,
836
+ 36,
837
+ 37,
838
+ 38,
839
+ 39
840
+ ],
841
+ "modality_keys": [
842
+ "left_wrist_eef",
843
+ "right_wrist_eef",
844
+ "left_gripper_pos",
845
+ "right_gripper_pos",
846
+ "left_joint_pos",
847
+ "right_joint_pos"
848
+ ],
849
+ "sin_cos_embedding_keys": null,
850
+ "mean_std_embedding_keys": null,
851
+ "action_configs": [
852
+ {
853
+ "rep": "RELATIVE",
854
+ "type": "EEF",
855
+ "format": "XYZ_ROT6D",
856
+ "state_key": "left_wrist_eef"
857
+ },
858
+ {
859
+ "rep": "RELATIVE",
860
+ "type": "EEF",
861
+ "format": "XYZ_ROT6D",
862
+ "state_key": "right_wrist_eef"
863
+ },
864
+ {
865
+ "rep": "ABSOLUTE",
866
+ "type": "NON_EEF",
867
+ "format": "DEFAULT",
868
+ "state_key": "left_gripper_pos"
869
+ },
870
+ {
871
+ "rep": "ABSOLUTE",
872
+ "type": "NON_EEF",
873
+ "format": "DEFAULT",
874
+ "state_key": "right_gripper_pos"
875
+ },
876
+ {
877
+ "rep": "RELATIVE",
878
+ "type": "NON_EEF",
879
+ "format": "DEFAULT",
880
+ "state_key": "left_joint_pos"
881
+ },
882
+ {
883
+ "rep": "RELATIVE",
884
+ "type": "NON_EEF",
885
+ "format": "DEFAULT",
886
+ "state_key": "right_joint_pos"
887
+ }
888
+ ]
889
+ },
890
+ "language": {
891
+ "delta_indices": [
892
+ 0
893
+ ],
894
+ "modality_keys": [
895
+ "annotation.sub_task"
896
+ ],
897
+ "sin_cos_embedding_keys": null,
898
+ "mean_std_embedding_keys": null,
899
+ "action_configs": null
900
+ }
901
+ },
902
+ "oxe_droid_relative_eef_relative_joint": {
903
+ "video": {
904
+ "delta_indices": [
905
+ -15,
906
+ 0
907
+ ],
908
+ "modality_keys": [
909
+ "exterior_image_1_left",
910
+ "wrist_image_left"
911
+ ],
912
+ "sin_cos_embedding_keys": null,
913
+ "mean_std_embedding_keys": null,
914
+ "action_configs": null
915
+ },
916
+ "state": {
917
+ "delta_indices": [
918
+ 0
919
+ ],
920
+ "modality_keys": [
921
+ "eef_9d",
922
+ "gripper_position",
923
+ "joint_position"
924
+ ],
925
+ "sin_cos_embedding_keys": null,
926
+ "mean_std_embedding_keys": null,
927
+ "action_configs": null
928
+ },
929
+ "action": {
930
+ "delta_indices": [
931
+ 0,
932
+ 1,
933
+ 2,
934
+ 3,
935
+ 4,
936
+ 5,
937
+ 6,
938
+ 7,
939
+ 8,
940
+ 9,
941
+ 10,
942
+ 11,
943
+ 12,
944
+ 13,
945
+ 14,
946
+ 15,
947
+ 16,
948
+ 17,
949
+ 18,
950
+ 19,
951
+ 20,
952
+ 21,
953
+ 22,
954
+ 23,
955
+ 24,
956
+ 25,
957
+ 26,
958
+ 27,
959
+ 28,
960
+ 29,
961
+ 30,
962
+ 31,
963
+ 32,
964
+ 33,
965
+ 34,
966
+ 35,
967
+ 36,
968
+ 37,
969
+ 38,
970
+ 39
971
+ ],
972
+ "modality_keys": [
973
+ "eef_9d",
974
+ "gripper_position",
975
+ "joint_position"
976
+ ],
977
+ "sin_cos_embedding_keys": null,
978
+ "mean_std_embedding_keys": null,
979
+ "action_configs": [
980
+ {
981
+ "rep": "RELATIVE",
982
+ "type": "EEF",
983
+ "format": "XYZ_ROT6D",
984
+ "state_key": "eef_9d"
985
+ },
986
+ {
987
+ "rep": "ABSOLUTE",
988
+ "type": "NON_EEF",
989
+ "format": "DEFAULT",
990
+ "state_key": "gripper_position"
991
+ },
992
+ {
993
+ "rep": "RELATIVE",
994
+ "type": "NON_EEF",
995
+ "format": "DEFAULT",
996
+ "state_key": "joint_position"
997
+ }
998
+ ]
999
+ },
1000
+ "language": {
1001
+ "delta_indices": [
1002
+ 0
1003
+ ],
1004
+ "modality_keys": [
1005
+ "annotation.language.language_instruction"
1006
+ ],
1007
+ "sin_cos_embedding_keys": null,
1008
+ "mean_std_embedding_keys": null,
1009
+ "action_configs": null
1010
+ }
1011
+ },
1012
+ "new_embodiment": {
1013
+ "video": {
1014
+ "delta_indices": [
1015
+ 0
1016
+ ],
1017
+ "modality_keys": [
1018
+ "high",
1019
+ "left_wrist_view",
1020
+ "right_wrist_view"
1021
+ ],
1022
+ "sin_cos_embedding_keys": null,
1023
+ "mean_std_embedding_keys": null,
1024
+ "action_configs": null
1025
+ },
1026
+ "state": {
1027
+ "delta_indices": [
1028
+ 0
1029
+ ],
1030
+ "modality_keys": [
1031
+ "left_arm",
1032
+ "right_arm",
1033
+ "left_hand",
1034
+ "right_hand"
1035
+ ],
1036
+ "sin_cos_embedding_keys": null,
1037
+ "mean_std_embedding_keys": null,
1038
+ "action_configs": null
1039
+ },
1040
+ "action": {
1041
+ "delta_indices": [
1042
+ 0,
1043
+ 1,
1044
+ 2,
1045
+ 3,
1046
+ 4,
1047
+ 5,
1048
+ 6,
1049
+ 7,
1050
+ 8,
1051
+ 9,
1052
+ 10,
1053
+ 11,
1054
+ 12,
1055
+ 13,
1056
+ 14,
1057
+ 15,
1058
+ 16,
1059
+ 17,
1060
+ 18,
1061
+ 19,
1062
+ 20,
1063
+ 21,
1064
+ 22,
1065
+ 23,
1066
+ 24,
1067
+ 25,
1068
+ 26,
1069
+ 27,
1070
+ 28,
1071
+ 29,
1072
+ 30,
1073
+ 31
1074
+ ],
1075
+ "modality_keys": [
1076
+ "left_arm",
1077
+ "right_arm",
1078
+ "left_hand",
1079
+ "right_hand"
1080
+ ],
1081
+ "sin_cos_embedding_keys": null,
1082
+ "mean_std_embedding_keys": null,
1083
+ "action_configs": [
1084
+ {
1085
+ "rep": "ABSOLUTE",
1086
+ "type": "NON_EEF",
1087
+ "format": "DEFAULT",
1088
+ "state_key": null
1089
+ },
1090
+ {
1091
+ "rep": "ABSOLUTE",
1092
+ "type": "NON_EEF",
1093
+ "format": "DEFAULT",
1094
+ "state_key": null
1095
+ },
1096
+ {
1097
+ "rep": "ABSOLUTE",
1098
+ "type": "NON_EEF",
1099
+ "format": "DEFAULT",
1100
+ "state_key": null
1101
+ },
1102
+ {
1103
+ "rep": "ABSOLUTE",
1104
+ "type": "NON_EEF",
1105
+ "format": "DEFAULT",
1106
+ "state_key": null
1107
+ }
1108
+ ]
1109
+ },
1110
+ "language": {
1111
+ "delta_indices": [
1112
+ 0
1113
+ ],
1114
+ "modality_keys": [
1115
+ "annotation.human.task_description"
1116
+ ],
1117
+ "sin_cos_embedding_keys": null,
1118
+ "mean_std_embedding_keys": null,
1119
+ "action_configs": null
1120
+ }
1121
+ }
1122
+ },
1123
+ "image_crop_size": [
1124
+ 230,
1125
+ 230
1126
+ ],
1127
+ "image_target_size": [
1128
+ 256,
1129
+ 256
1130
+ ],
1131
+ "use_albumentations": true,
1132
+ "random_rotation_angle": 0,
1133
+ "color_jitter_params": {
1134
+ "brightness": 0.3,
1135
+ "contrast": 0.4,
1136
+ "saturation": 0.5,
1137
+ "hue": 0.08
1138
+ },
1139
+ "shortest_image_edge": 256,
1140
+ "crop_fraction": 0.95,
1141
+ "letter_box_transform": false,
1142
+ "model_name": "nvidia/Cosmos-Reason2-2B",
1143
+ "model_type": "qwen",
1144
+ "formalize_language": true,
1145
+ "max_state_dim": 132,
1146
+ "max_action_dim": 132,
1147
+ "max_action_horizon": 40,
1148
+ "use_percentiles": true,
1149
+ "use_mean_std": false,
1150
+ "clip_outliers": true,
1151
+ "apply_sincos_state_encoding": false,
1152
+ "use_relative_action": true,
1153
+ "exclude_state": false,
1154
+ "state_dropout_prob": 0.0
1155
+ }
1156
+ }
SFT/pack_agx_mimicgen_sft_fixData-1/checkpoint-40000/statistics.json ADDED
The diff for this file is too large to render. See raw diff