Hugging Face's logo

Grigorij
/
PaP_objects_2_groot

Safetensors
Gr00tN1d7
Model card Files
xet
 Community
Grigorij commited on
Commit
5e505c4
·
verified ·
1 Parent(s): 1b5ce9e

Upload folder using huggingface_hub

Browse files
Files changed (18) hide show
  1. config.json +88 -0
  2. embodiment_id.json +57 -0
  3. experiment_cfg/conf.yaml +215 -0
  4. experiment_cfg/config.yaml +213 -0
  5. experiment_cfg/dataset_statistics.json +824 -0
  6. experiment_cfg/final_model_config.json +56 -0
  7. experiment_cfg/final_processor_config.json +0 -0
  8. model-00001-of-00003.safetensors +3 -0
  9. model-00002-of-00003.safetensors +3 -0
  10. model-00003-of-00003.safetensors +3 -0
  11. model.safetensors.index.json +0 -0
  12. processor_config.json +1123 -0
  13. rng_state.pth +3 -0
  14. scheduler.pt +3 -0
  15. statistics.json +0 -0
  16. trainer_state.json +0 -0
  17. training_args.bin +3 -0
  18. wandb_config.json +1 -0
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": "float32",
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": 0.2,
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": false,
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
+ }
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
+ }
experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,215 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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: false
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: null
27
+ use_albumentations_transforms: true
28
+ letter_box_transform: false
29
+ extra_augmentation_config: null
30
+ formalize_language: true
31
+ apply_sincos_state_encoding: false
32
+ use_percentiles: true
33
+ use_relative_action: true
34
+ max_state_dim: 132
35
+ max_action_dim: 132
36
+ action_horizon: 40
37
+ hidden_size: 1024
38
+ input_embedding_dim: 1536
39
+ state_history_length: 1
40
+ add_pos_embed: true
41
+ attn_dropout: 0.2
42
+ use_vlln: true
43
+ max_seq_len: 1024
44
+ use_alternate_vl_dit: true
45
+ attend_text_every_n_blocks: 2
46
+ diffusion_model_cfg:
47
+ positional_embeddings: null
48
+ num_layers: 16
49
+ num_attention_heads: 32
50
+ attention_head_dim: 48
51
+ norm_type: ada_norm
52
+ dropout: 0.2
53
+ final_dropout: true
54
+ output_dim: 1024
55
+ interleave_self_attention: true
56
+ num_inference_timesteps: 4
57
+ noise_beta_alpha: 1.5
58
+ noise_beta_beta: 1.0
59
+ noise_s: 0.999
60
+ num_timestep_buckets: 1000
61
+ tune_projector: true
62
+ tune_diffusion_model: true
63
+ tune_vlln: true
64
+ state_dropout_prob: 0.2
65
+ exclude_state: false
66
+ use_mean_std: false
67
+ max_num_embodiments: 32
68
+ data:
69
+ datasets:
70
+ - dataset_paths:
71
+ - /workspace/data/Grigorij/PaP_objects_2_sum
72
+ embodiment_tag: new_embodiment
73
+ mix_ratio: 1.0
74
+ dataset_type: physical_embodiment
75
+ val_dataset_path: null
76
+ modality_configs:
77
+ new_embodiment:
78
+ video:
79
+ delta_indices:
80
+ - 0
81
+ modality_keys:
82
+ - front
83
+ - wrist
84
+ sin_cos_embedding_keys: null
85
+ mean_std_embedding_keys: null
86
+ action_configs: null
87
+ state:
88
+ delta_indices:
89
+ - 0
90
+ modality_keys:
91
+ - single_arm
92
+ - gripper
93
+ sin_cos_embedding_keys: null
94
+ mean_std_embedding_keys: null
95
+ action_configs: null
96
+ action:
97
+ delta_indices:
98
+ - 0
99
+ - 1
100
+ - 2
101
+ - 3
102
+ - 4
103
+ - 5
104
+ - 6
105
+ - 7
106
+ - 8
107
+ - 9
108
+ - 10
109
+ - 11
110
+ - 12
111
+ - 13
112
+ - 14
113
+ - 15
114
+ modality_keys:
115
+ - single_arm
116
+ - gripper
117
+ sin_cos_embedding_keys: null
118
+ mean_std_embedding_keys: null
119
+ action_configs:
120
+ - rep: RELATIVE
121
+ type: NON_EEF
122
+ format: DEFAULT
123
+ state_key: null
124
+ - rep: ABSOLUTE
125
+ type: NON_EEF
126
+ format: DEFAULT
127
+ state_key: null
128
+ language:
129
+ delta_indices:
130
+ - 0
131
+ modality_keys:
132
+ - annotation.human.task_description
133
+ sin_cos_embedding_keys: null
134
+ mean_std_embedding_keys: null
135
+ action_configs: null
136
+ download_cache: false
137
+ shard_size: 1024
138
+ episode_sampling_rate: 0.1
139
+ num_shards_per_epoch: 100000
140
+ ds_weights_alpha: null
141
+ override_pretraining_statistics: true
142
+ mode: single_turn
143
+ random_chop: 0.0
144
+ mock_dataset_mode: false
145
+ shuffle: true
146
+ seed: 42
147
+ multiprocessing_context: fork
148
+ allow_padding: false
149
+ subsample_ratio: 1.0
150
+ image_crop_size:
151
+ - 244
152
+ - 244
153
+ image_target_size:
154
+ - 224
155
+ - 224
156
+ training:
157
+ output_dir: /workspace/outputs/groot_objects2
158
+ experiment_name: null
159
+ max_steps: 20000
160
+ global_batch_size: 64
161
+ per_gpu_batch_size: null
162
+ gradient_accumulation_steps: 1
163
+ learning_rate: 0.0001
164
+ lr_scheduler_type: cosine
165
+ weight_decay: 1.0e-05
166
+ warmup_ratio: 0.05
167
+ warmup_steps: 0
168
+ max_grad_norm: 1.0
169
+ optim: adamw_torch
170
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
171
+ skip_weight_loading: false
172
+ tf32: true
173
+ fp16: false
174
+ bf16: true
175
+ eval_bf16: true
176
+ logging_steps: 10
177
+ save_steps: 5000
178
+ save_total_limit: 1
179
+ save_vl_model: false
180
+ save_only_model: false
181
+ resume_from_checkpoint: false
182
+ upload_checkpoints: false
183
+ upload_every: 1000
184
+ upload_last_n_checkpoints: 5
185
+ max_concurrent_uploads: 2
186
+ eval_strategy: 'no'
187
+ eval_steps: 500
188
+ eval_set_split_ratio: 0.1
189
+ eval_batch_size: 2
190
+ save_best_eval_metric_name: ''
191
+ save_best_eval_metric_greater_is_better: true
192
+ deepspeed_stage: 2
193
+ gradient_checkpointing: false
194
+ transformers_trust_remote_code: true
195
+ transformers_local_files_only: false
196
+ transformers_cache_dir: null
197
+ transformers_access_token: null
198
+ use_ddp: false
199
+ ddp_bucket_cap_mb: 100
200
+ num_gpus: 1
201
+ dataloader_num_workers: 8
202
+ remove_unused_columns: false
203
+ use_wandb: false
204
+ wandb_project: finetune-gr00t-n1d7
205
+ enable_profiling: false
206
+ max_retries: 3
207
+ assert_loss_less_than: null
208
+ add_rl_callback: false
209
+ enable_open_loop_eval: false
210
+ open_loop_eval_traj_ids:
211
+ - 0
212
+ open_loop_eval_steps_per_traj: 100
213
+ open_loop_eval_plot_indices: null
214
+ max_steps: 20000
215
+ save_steps: 5000
experiment_cfg/config.yaml ADDED
@@ -0,0 +1,213 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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: false
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: null
27
+ use_albumentations_transforms: true
28
+ letter_box_transform: false
29
+ extra_augmentation_config: null
30
+ formalize_language: true
31
+ apply_sincos_state_encoding: false
32
+ use_percentiles: true
33
+ use_relative_action: true
34
+ max_state_dim: 132
35
+ max_action_dim: 132
36
+ action_horizon: 40
37
+ hidden_size: 1024
38
+ input_embedding_dim: 1536
39
+ state_history_length: 1
40
+ add_pos_embed: true
41
+ attn_dropout: 0.2
42
+ use_vlln: true
43
+ max_seq_len: 1024
44
+ use_alternate_vl_dit: true
45
+ attend_text_every_n_blocks: 2
46
+ diffusion_model_cfg:
47
+ positional_embeddings: null
48
+ num_layers: 16
49
+ num_attention_heads: 32
50
+ attention_head_dim: 48
51
+ norm_type: ada_norm
52
+ dropout: 0.2
53
+ final_dropout: true
54
+ output_dim: 1024
55
+ interleave_self_attention: true
56
+ num_inference_timesteps: 4
57
+ noise_beta_alpha: 1.5
58
+ noise_beta_beta: 1.0
59
+ noise_s: 0.999
60
+ num_timestep_buckets: 1000
61
+ tune_projector: true
62
+ tune_diffusion_model: true
63
+ tune_vlln: true
64
+ state_dropout_prob: 0.2
65
+ exclude_state: false
66
+ use_mean_std: false
67
+ max_num_embodiments: 32
68
+ data:
69
+ datasets:
70
+ - dataset_paths:
71
+ - /workspace/data/Grigorij/PaP_objects_2_sum
72
+ embodiment_tag: new_embodiment
73
+ mix_ratio: 1.0
74
+ dataset_type: physical_embodiment
75
+ val_dataset_path: null
76
+ modality_configs:
77
+ new_embodiment:
78
+ video:
79
+ delta_indices:
80
+ - 0
81
+ modality_keys:
82
+ - front
83
+ - wrist
84
+ sin_cos_embedding_keys: null
85
+ mean_std_embedding_keys: null
86
+ action_configs: null
87
+ state:
88
+ delta_indices:
89
+ - 0
90
+ modality_keys:
91
+ - single_arm
92
+ - gripper
93
+ sin_cos_embedding_keys: null
94
+ mean_std_embedding_keys: null
95
+ action_configs: null
96
+ action:
97
+ delta_indices:
98
+ - 0
99
+ - 1
100
+ - 2
101
+ - 3
102
+ - 4
103
+ - 5
104
+ - 6
105
+ - 7
106
+ - 8
107
+ - 9
108
+ - 10
109
+ - 11
110
+ - 12
111
+ - 13
112
+ - 14
113
+ - 15
114
+ modality_keys:
115
+ - single_arm
116
+ - gripper
117
+ sin_cos_embedding_keys: null
118
+ mean_std_embedding_keys: null
119
+ action_configs:
120
+ - rep: relative
121
+ type: non_eef
122
+ format: default
123
+ state_key: null
124
+ - rep: absolute
125
+ type: non_eef
126
+ format: default
127
+ state_key: null
128
+ language:
129
+ delta_indices:
130
+ - 0
131
+ modality_keys:
132
+ - annotation.human.task_description
133
+ sin_cos_embedding_keys: null
134
+ mean_std_embedding_keys: null
135
+ action_configs: null
136
+ download_cache: false
137
+ shard_size: 1024
138
+ episode_sampling_rate: 0.1
139
+ num_shards_per_epoch: 100000
140
+ ds_weights_alpha: null
141
+ override_pretraining_statistics: true
142
+ mode: single_turn
143
+ random_chop: 0.0
144
+ mock_dataset_mode: false
145
+ shuffle: true
146
+ seed: 42
147
+ multiprocessing_context: fork
148
+ allow_padding: false
149
+ subsample_ratio: 1.0
150
+ image_crop_size:
151
+ - 244
152
+ - 244
153
+ image_target_size:
154
+ - 224
155
+ - 224
156
+ training:
157
+ output_dir: /workspace/outputs/groot_objects2
158
+ experiment_name: null
159
+ max_steps: 20000
160
+ global_batch_size: 64
161
+ per_gpu_batch_size: null
162
+ gradient_accumulation_steps: 1
163
+ learning_rate: 0.0001
164
+ lr_scheduler_type: cosine
165
+ weight_decay: 1.0e-05
166
+ warmup_ratio: 0.05
167
+ warmup_steps: 0
168
+ max_grad_norm: 1.0
169
+ optim: adamw_torch
170
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
171
+ skip_weight_loading: false
172
+ tf32: true
173
+ fp16: false
174
+ bf16: true
175
+ eval_bf16: true
176
+ logging_steps: 10
177
+ save_steps: 5000
178
+ save_total_limit: 1
179
+ save_vl_model: false
180
+ save_only_model: false
181
+ resume_from_checkpoint: false
182
+ upload_checkpoints: false
183
+ upload_every: 1000
184
+ upload_last_n_checkpoints: 5
185
+ max_concurrent_uploads: 2
186
+ eval_strategy: 'no'
187
+ eval_steps: 500
188
+ eval_set_split_ratio: 0.1
189
+ eval_batch_size: 2
190
+ save_best_eval_metric_name: ''
191
+ save_best_eval_metric_greater_is_better: true
192
+ deepspeed_stage: 2
193
+ gradient_checkpointing: false
194
+ transformers_trust_remote_code: true
195
+ transformers_local_files_only: false
196
+ transformers_cache_dir: null
197
+ transformers_access_token: null
198
+ use_ddp: false
199
+ ddp_bucket_cap_mb: 100
200
+ num_gpus: 1
201
+ dataloader_num_workers: 8
202
+ remove_unused_columns: false
203
+ use_wandb: false
204
+ wandb_project: finetune-gr00t-n1d7
205
+ enable_profiling: false
206
+ max_retries: 3
207
+ assert_loss_less_than: null
208
+ add_rl_callback: false
209
+ enable_open_loop_eval: false
210
+ open_loop_eval_traj_ids:
211
+ - 0
212
+ open_loop_eval_steps_per_traj: 100
213
+ open_loop_eval_plot_indices: null
experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,824 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "single_arm": {
5
+ "min": [
6
+ -61.27472686767578,
7
+ -103.95604705810547,
8
+ -41.846153259277344,
9
+ 12.747252464294434,
10
+ -160.21978759765625
11
+ ],
12
+ "max": [
13
+ 48.17582321166992,
14
+ 50.153846740722656,
15
+ 97.4945068359375,
16
+ 96.70330047607422,
17
+ -6.989010810852051
18
+ ],
19
+ "mean": [
20
+ -1.9332756996154785,
21
+ -54.635982513427734,
22
+ 58.85205841064453,
23
+ 64.550048828125,
24
+ -88.13927459716797
25
+ ],
26
+ "std": [
27
+ 21.783662796020508,
28
+ 45.6338005065918,
29
+ 37.80471420288086,
30
+ 11.509209632873535,
31
+ 20.170608520507812
32
+ ],
33
+ "q01": [
34
+ -54.94505310058594,
35
+ -102.54945373535156,
36
+ -21.186813354492188,
37
+ 36.39560317993164,
38
+ -140.87911987304688
39
+ ],
40
+ "q99": [
41
+ 42.72527313232422,
42
+ 35.03296661376953,
43
+ 97.05494689941406,
44
+ 89.23076629638672,
45
+ -33.537580680846986
46
+ ]
47
+ },
48
+ "gripper": {
49
+ "min": [
50
+ 1.9512195587158203
51
+ ],
52
+ "max": [
53
+ 77.15447235107422
54
+ ],
55
+ "mean": [
56
+ 11.448285102844238
57
+ ],
58
+ "std": [
59
+ 11.980111122131348
60
+ ],
61
+ "q01": [
62
+ 2.032520294189453
63
+ ],
64
+ "q99": [
65
+ 49.43089294433594
66
+ ]
67
+ }
68
+ },
69
+ "action": {
70
+ "single_arm": {
71
+ "min": [
72
+ -61.758243560791016,
73
+ -103.9120864868164,
74
+ -43.69230651855469,
75
+ 11.296703338623047,
76
+ -160.57142639160156
77
+ ],
78
+ "max": [
79
+ 48.30769348144531,
80
+ 52.483516693115234,
81
+ 97.75823974609375,
82
+ 98.24176025390625,
83
+ -5.758241653442383
84
+ ],
85
+ "mean": [
86
+ -1.9997183084487915,
87
+ -55.236328125,
88
+ 57.90788269042969,
89
+ 64.93882751464844,
90
+ -88.12054443359375
91
+ ],
92
+ "std": [
93
+ 21.789033889770508,
94
+ 45.10258483886719,
95
+ 38.51021957397461,
96
+ 12.326508522033691,
97
+ 20.233089447021484
98
+ ],
99
+ "q01": [
100
+ -54.989009857177734,
101
+ -102.4175796508789,
102
+ -22.94505500793457,
103
+ 36.43955993652344,
104
+ -140.7920930480957
105
+ ],
106
+ "q99": [
107
+ 42.68132019042969,
108
+ 33.93494586944598,
109
+ 97.05494689941406,
110
+ 89.45054626464844,
111
+ -32.92307662963867
112
+ ]
113
+ },
114
+ "gripper": {
115
+ "min": [
116
+ 0.24793387949466705
117
+ ],
118
+ "max": [
119
+ 77.52066040039062
120
+ ],
121
+ "mean": [
122
+ 10.434662818908691
123
+ ],
124
+ "std": [
125
+ 12.608206748962402
126
+ ],
127
+ "q01": [
128
+ 1.6528925895690918
129
+ ],
130
+ "q99": [
131
+ 49.421485900878906
132
+ ]
133
+ }
134
+ },
135
+ "relative_action": {
136
+ "single_arm": {
137
+ "min": [
138
+ [
139
+ -14.19780158996582,
140
+ -14.285717010498047,
141
+ -16.70330047607422,
142
+ -15.956043243408203,
143
+ -14.945053100585938
144
+ ],
145
+ [
146
+ -17.80219841003418,
147
+ -17.89011001586914,
148
+ -17.846153259277344,
149
+ -19.824176788330078,
150
+ -15.296707153320312
151
+ ],
152
+ [
153
+ -21.582416534423828,
154
+ -21.318683624267578,
155
+ -19.868133544921875,
156
+ -23.604393005371094,
157
+ -17.582420349121094
158
+ ],
159
+ [
160
+ -25.186813354492188,
161
+ -24.659343719482422,
162
+ -22.681320190429688,
163
+ -27.208789825439453,
164
+ -18.813186645507812
165
+ ],
166
+ [
167
+ -28.615385055541992,
168
+ -28.0,
169
+ -25.75823974609375,
170
+ -30.72527313232422,
171
+ -21.362632751464844
172
+ ],
173
+ [
174
+ -32.0439567565918,
175
+ -31.25274658203125,
176
+ -28.835163116455078,
177
+ -33.80219650268555,
178
+ -23.912086486816406
179
+ ],
180
+ [
181
+ -35.47252655029297,
182
+ -34.41758346557617,
183
+ -31.736263275146484,
184
+ -36.79120635986328,
185
+ -26.637359619140625
186
+ ],
187
+ [
188
+ -38.72527313232422,
189
+ -37.582420349121094,
190
+ -34.4615364074707,
191
+ -39.07692337036133,
192
+ -29.186813354492188
193
+ ],
194
+ [
195
+ -41.80219650268555,
196
+ -40.74725341796875,
197
+ -36.83516311645508,
198
+ -41.18681335449219,
199
+ -31.560440063476562
200
+ ],
201
+ [
202
+ -44.79120635986328,
203
+ -43.82417297363281,
204
+ -39.29670333862305,
205
+ -42.94505310058594,
206
+ -33.93406677246094
207
+ ],
208
+ [
209
+ -47.78022003173828,
210
+ -46.72527313232422,
211
+ -41.40659713745117,
212
+ -44.087913513183594,
213
+ -36.30769348144531
214
+ ],
215
+ [
216
+ -50.5054931640625,
217
+ -49.626373291015625,
218
+ -43.340660095214844,
219
+ -45.05494689941406,
220
+ -38.5054931640625
221
+ ],
222
+ [
223
+ -53.142860412597656,
224
+ -52.52747344970703,
225
+ -45.098899841308594,
226
+ -47.252750396728516,
227
+ -40.79121017456055
228
+ ],
229
+ [
230
+ -55.78022003173828,
231
+ -55.340660095214844,
232
+ -46.769229888916016,
233
+ -49.274723052978516,
234
+ -42.989009857177734
235
+ ],
236
+ [
237
+ -58.329673767089844,
238
+ -58.06593322753906,
239
+ -48.0,
240
+ -51.20878601074219,
241
+ -45.010990142822266
242
+ ],
243
+ [
244
+ -60.879119873046875,
245
+ -60.615386962890625,
246
+ -49.230770111083984,
247
+ -52.879119873046875,
248
+ -47.032962799072266
249
+ ]
250
+ ],
251
+ "max": [
252
+ [
253
+ 13.582417488098145,
254
+ 16.131866455078125,
255
+ 15.38461685180664,
256
+ 31.076927185058594,
257
+ 20.835166931152344
258
+ ],
259
+ [
260
+ 14.725275039672852,
261
+ 20.26373291015625,
262
+ 19.692306518554688,
263
+ 31.076927185058594,
264
+ 20.835166931152344
265
+ ],
266
+ [
267
+ 15.868131637573242,
268
+ 23.78022003173828,
269
+ 24.0,
270
+ 31.076927185058594,
271
+ 21.71428680419922
272
+ ],
273
+ [
274
+ 17.89011001586914,
275
+ 27.208789825439453,
276
+ 28.21977996826172,
277
+ 30.813186645507812,
278
+ 25.318679809570312
279
+ ],
280
+ [
281
+ 20.439559936523438,
282
+ 30.549449920654297,
283
+ 32.52747344970703,
284
+ 31.86812973022461,
285
+ 28.395606994628906
286
+ ],
287
+ [
288
+ 22.813186645507812,
289
+ 33.89011001586914,
290
+ 36.74725341796875,
291
+ 33.80219650268555,
292
+ 31.47252655029297
293
+ ],
294
+ [
295
+ 25.186811447143555,
296
+ 36.87912368774414,
297
+ 40.79120635986328,
298
+ 35.5604362487793,
299
+ 34.19780731201172
300
+ ],
301
+ [
302
+ 27.4725284576416,
303
+ 39.69231033325195,
304
+ 44.747249603271484,
305
+ 37.0549430847168,
306
+ 36.74725341796875
307
+ ],
308
+ [
309
+ 29.670330047607422,
310
+ 42.41758346557617,
311
+ 48.4395637512207,
312
+ 37.75823974609375,
313
+ 39.120880126953125
314
+ ],
315
+ [
316
+ 31.868133544921875,
317
+ 45.142860412597656,
318
+ 52.30768966674805,
319
+ 38.19779968261719,
320
+ 41.31867980957031
321
+ ],
322
+ [
323
+ 34.06593704223633,
324
+ 47.69230651855469,
325
+ 56.000003814697266,
326
+ 39.252750396728516,
327
+ 43.16484069824219
328
+ ],
329
+ [
330
+ 36.263736724853516,
331
+ 50.15384292602539,
332
+ 59.60439682006836,
333
+ 40.483516693115234,
334
+ 44.659339904785156
335
+ ],
336
+ [
337
+ 38.28571319580078,
338
+ 52.43955612182617,
339
+ 63.20878982543945,
340
+ 41.36263656616211,
341
+ 45.890113830566406
342
+ ],
343
+ [
344
+ 40.131866455078125,
345
+ 54.637359619140625,
346
+ 66.72527313232422,
347
+ 42.5054931640625,
348
+ 46.85714340209961
349
+ ],
350
+ [
351
+ 41.97801971435547,
352
+ 56.57142639160156,
353
+ 70.24176025390625,
354
+ 43.120880126953125,
355
+ 47.56044006347656
356
+ ],
357
+ [
358
+ 43.56044006347656,
359
+ 58.32966995239258,
360
+ 73.67033386230469,
361
+ 43.472530364990234,
362
+ 48.175819396972656
363
+ ]
364
+ ],
365
+ "mean": [
366
+ [
367
+ -0.06283990293741226,
368
+ -0.613447368144989,
369
+ -0.9790156483650208,
370
+ 0.3510562479496002,
371
+ 0.020385928452014923
372
+ ],
373
+ [
374
+ -0.06374113261699677,
375
+ -0.6129622459411621,
376
+ -0.9795311093330383,
377
+ 0.3488035500049591,
378
+ 0.021142004057765007
379
+ ],
380
+ [
381
+ -0.06469117105007172,
382
+ -0.6124843955039978,
383
+ -0.9800305962562561,
384
+ 0.34645339846611023,
385
+ 0.021941259503364563
386
+ ],
387
+ [
388
+ -0.06568773090839386,
389
+ -0.6120145916938782,
390
+ -0.9805121421813965,
391
+ 0.34402045607566833,
392
+ 0.02281254716217518
393
+ ],
394
+ [
395
+ -0.06673634052276611,
396
+ -0.6115391850471497,
397
+ -0.9809795618057251,
398
+ 0.34157782793045044,
399
+ 0.02373875305056572
400
+ ],
401
+ [
402
+ -0.0678638443350792,
403
+ -0.6110692620277405,
404
+ -0.98143470287323,
405
+ 0.33918264508247375,
406
+ 0.024717001244425774
407
+ ],
408
+ [
409
+ -0.06907179206609726,
410
+ -0.6106093525886536,
411
+ -0.9818698167800903,
412
+ 0.33685413002967834,
413
+ 0.025756508111953735
414
+ ],
415
+ [
416
+ -0.07036781311035156,
417
+ -0.6101579666137695,
418
+ -0.9822930097579956,
419
+ 0.334551066160202,
420
+ 0.026878563687205315
421
+ ],
422
+ [
423
+ -0.07173533737659454,
424
+ -0.6097102761268616,
425
+ -0.9826939702033997,
426
+ 0.33225277066230774,
427
+ 0.02810450829565525
428
+ ],
429
+ [
430
+ -0.07317031174898148,
431
+ -0.609282374382019,
432
+ -0.983081042766571,
433
+ 0.3299757242202759,
434
+ 0.029432915151119232
435
+ ],
436
+ [
437
+ -0.07466857135295868,
438
+ -0.6089041829109192,
439
+ -0.9834362268447876,
440
+ 0.3277260363101959,
441
+ 0.030859487131237984
442
+ ],
443
+ [
444
+ -0.07625292986631393,
445
+ -0.6086094975471497,
446
+ -0.9837437868118286,
447
+ 0.32552242279052734,
448
+ 0.03233977034687996
449
+ ],
450
+ [
451
+ -0.0779147669672966,
452
+ -0.6084027886390686,
453
+ -0.9839838743209839,
454
+ 0.3233998119831085,
455
+ 0.033854592591524124
456
+ ],
457
+ [
458
+ -0.07964973151683807,
459
+ -0.6083223819732666,
460
+ -0.9841431975364685,
461
+ 0.32140839099884033,
462
+ 0.03540864959359169
463
+ ],
464
+ [
465
+ -0.08154365420341492,
466
+ -0.6084256768226624,
467
+ -0.9842016100883484,
468
+ 0.31956297159194946,
469
+ 0.03702140226960182
470
+ ],
471
+ [
472
+ -0.08361750841140747,
473
+ -0.6087486147880554,
474
+ -0.9840916991233826,
475
+ 0.31787416338920593,
476
+ 0.038714464753866196
477
+ ]
478
+ ],
479
+ "std": [
480
+ [
481
+ 2.0244054794311523,
482
+ 3.012411594390869,
483
+ 2.658233642578125,
484
+ 2.272719144821167,
485
+ 1.3880157470703125
486
+ ],
487
+ [
488
+ 2.5934579372406006,
489
+ 3.7631289958953857,
490
+ 3.264678478240967,
491
+ 2.5601766109466553,
492
+ 1.744330644607544
493
+ ],
494
+ [
495
+ 3.1596884727478027,
496
+ 4.515340805053711,
497
+ 3.878251075744629,
498
+ 2.8616604804992676,
499
+ 2.0934927463531494
500
+ ],
501
+ [
502
+ 3.7210988998413086,
503
+ 5.264068126678467,
504
+ 4.492758750915527,
505
+ 3.166059970855713,
506
+ 2.4330039024353027
507
+ ],
508
+ [
509
+ 4.2769598960876465,
510
+ 6.007012844085693,
511
+ 5.104133605957031,
512
+ 3.4669997692108154,
513
+ 2.7621772289276123
514
+ ],
515
+ [
516
+ 4.82646369934082,
517
+ 6.742584705352783,
518
+ 5.710707664489746,
519
+ 3.761524200439453,
520
+ 3.0808491706848145
521
+ ],
522
+ [
523
+ 5.369154930114746,
524
+ 7.469855308532715,
525
+ 6.311351776123047,
526
+ 4.047056674957275,
527
+ 3.3891119956970215
528
+ ],
529
+ [
530
+ 5.904894828796387,
531
+ 8.188711166381836,
532
+ 6.90587854385376,
533
+ 4.322752952575684,
534
+ 3.6876232624053955
535
+ ],
536
+ [
537
+ 6.433441162109375,
538
+ 8.898999214172363,
539
+ 7.493752479553223,
540
+ 4.587961673736572,
541
+ 3.9768242835998535
542
+ ],
543
+ [
544
+ 6.954634189605713,
545
+ 9.600430488586426,
546
+ 8.074893951416016,
547
+ 4.8429131507873535,
548
+ 4.257540702819824
549
+ ],
550
+ [
551
+ 7.46846342086792,
552
+ 10.293462753295898,
553
+ 8.649243354797363,
554
+ 5.087597846984863,
555
+ 4.530071258544922
556
+ ],
557
+ [
558
+ 7.9748711585998535,
559
+ 10.978036880493164,
560
+ 9.216875076293945,
561
+ 5.322606563568115,
562
+ 4.795181751251221
563
+ ],
564
+ [
565
+ 8.473917007446289,
566
+ 11.654413223266602,
567
+ 9.777687072753906,
568
+ 5.548206806182861,
569
+ 5.053039073944092
570
+ ],
571
+ [
572
+ 8.965482711791992,
573
+ 12.322737693786621,
574
+ 10.332100868225098,
575
+ 5.764930248260498,
576
+ 5.304286956787109
577
+ ],
578
+ [
579
+ 9.449603080749512,
580
+ 12.983320236206055,
581
+ 10.880204200744629,
582
+ 5.972500324249268,
583
+ 5.549320220947266
584
+ ],
585
+ [
586
+ 9.926215171813965,
587
+ 13.636313438415527,
588
+ 11.422407150268555,
589
+ 6.171347618103027,
590
+ 5.788440227508545
591
+ ]
592
+ ],
593
+ "q01": [
594
+ [
595
+ -7.076923370361328,
596
+ -9.186813354492188,
597
+ -8.175827026367188,
598
+ -5.230766296386719,
599
+ -4.659339904785156
600
+ ],
601
+ [
602
+ -9.186813354492188,
603
+ -11.384614944458008,
604
+ -9.934066772460938,
605
+ -6.5494537353515625,
606
+ -5.890113830566406
607
+ ],
608
+ [
609
+ -11.208791732788086,
610
+ -13.582416534423828,
611
+ -11.604400634765625,
612
+ -7.868133544921875,
613
+ -7.2536256408691395
614
+ ],
615
+ [
616
+ -13.31868124961853,
617
+ -15.780219564437866,
618
+ -13.36263656616211,
619
+ -9.098899841308594,
620
+ -8.439559936523438
621
+ ],
622
+ [
623
+ -15.340659141540527,
624
+ -17.89011001586914,
625
+ -14.945056915283203,
626
+ -10.24176025390625,
627
+ -9.539342727661133
628
+ ],
629
+ [
630
+ -17.36263656616211,
631
+ -20.0,
632
+ -16.615386962890625,
633
+ -11.296703338623047,
634
+ -10.813186645507812
635
+ ],
636
+ [
637
+ -19.384614944458008,
638
+ -22.10988998413086,
639
+ -18.24263931274414,
640
+ -12.351646423339844,
641
+ -11.868127861022948
642
+ ],
643
+ [
644
+ -21.318681716918945,
645
+ -24.21977996826172,
646
+ -19.78021816253662,
647
+ -13.230772056579589,
648
+ -12.923072814941406
649
+ ],
650
+ [
651
+ -23.340659160614013,
652
+ -26.241757411956787,
653
+ -21.186813354492188,
654
+ -14.10989047050476,
655
+ -13.890106201171875
656
+ ],
657
+ [
658
+ -25.27472496032715,
659
+ -28.175826091766357,
660
+ -22.681317348480224,
661
+ -15.076919555664062,
662
+ -14.769233703613281
663
+ ],
664
+ [
665
+ -27.032967586517334,
666
+ -30.109892864227294,
667
+ -24.04483386993408,
668
+ -15.78021816253662,
669
+ -15.693185577392578
670
+ ],
671
+ [
672
+ -29.011867332458497,
673
+ -32.0439587020874,
674
+ -25.494501152038573,
675
+ -16.52835250854492,
676
+ -16.615379333496094
677
+ ],
678
+ [
679
+ -30.81318473815918,
680
+ -33.97802165985107,
681
+ -26.72527313232422,
682
+ -17.143737602233887,
683
+ -17.494502124786376
684
+ ],
685
+ [
686
+ -32.57143020629883,
687
+ -35.82417297363281,
688
+ -28.176706085205076,
689
+ -17.802200317382812,
690
+ -18.28571708679199
691
+ ],
692
+ [
693
+ -34.32966995239258,
694
+ -37.58241653442383,
695
+ -29.450546264648438,
696
+ -18.329669952392578,
697
+ -19.164838829040526
698
+ ],
699
+ [
700
+ -36.087913513183594,
701
+ -39.42857360839844,
702
+ -30.681320190429688,
703
+ -18.945053100585938,
704
+ -19.956043319702147
705
+ ]
706
+ ],
707
+ "q99": [
708
+ [
709
+ 5.846153259277344,
710
+ 7.164833068847656,
711
+ 6.417579650878906,
712
+ 6.4615325927734375,
713
+ 4.835166931152344
714
+ ],
715
+ [
716
+ 7.516484026908874,
717
+ 9.274726867675781,
718
+ 8.615379333496094,
719
+ 6.989013671875,
720
+ 6.24176025390625
721
+ ],
722
+ [
723
+ 9.231647300720393,
724
+ 11.296705284118651,
725
+ 10.813186645507812,
726
+ 8.483516693115234,
727
+ 7.5604400634765625
728
+ ],
729
+ [
730
+ 10.85714340209961,
731
+ 13.318681755065917,
732
+ 12.923076629638672,
733
+ 9.890106201171875,
734
+ 8.791206359863281
735
+ ],
736
+ [
737
+ 12.43956184387207,
738
+ 15.340660095214844,
739
+ 15.0329656791687,
740
+ 11.208791770935058,
741
+ 9.934066772460938
742
+ ],
743
+ [
744
+ 14.0219793510437,
745
+ 17.362632751464844,
746
+ 17.054946899414062,
747
+ 12.35165023803711,
748
+ 11.076927185058594
749
+ ],
750
+ [
751
+ 15.516485214233398,
752
+ 19.29669952392578,
753
+ 19.076923370361328,
754
+ 13.715165863037289,
755
+ 12.131866455078125
756
+ ],
757
+ [
758
+ 17.05582408905047,
759
+ 21.054946899414062,
760
+ 21.098900814056396,
761
+ 14.725273132324219,
762
+ 13.098899841308594
763
+ ],
764
+ [
765
+ 18.505495071411133,
766
+ 22.90109447479248,
767
+ 23.07780429840106,
768
+ 15.780220031738281,
769
+ 13.978019714355469
770
+ ],
771
+ [
772
+ 19.824176788330078,
773
+ 24.61626407623309,
774
+ 25.14285659790039,
775
+ 16.923072814941406,
776
+ 14.857145347595214
777
+ ],
778
+ [
779
+ 21.142858505249023,
780
+ 26.28659225463885,
781
+ 27.164833068847656,
782
+ 17.802196502685547,
783
+ 15.781099090576351
784
+ ],
785
+ [
786
+ 22.54945182800293,
787
+ 27.95692611694354,
788
+ 28.96791244506854,
789
+ 18.681322135925292,
790
+ 16.52747344970703
791
+ ],
792
+ [
793
+ 23.78022003173828,
794
+ 29.49450496673584,
795
+ 30.814065704345882,
796
+ 19.472530364990234,
797
+ 17.49450496673584
798
+ ],
799
+ [
800
+ 25.098901748657227,
801
+ 31.076919555664062,
802
+ 32.61538314819336,
803
+ 20.175827026367188,
804
+ 18.1098881149292
805
+ ],
806
+ [
807
+ 26.329669952392578,
808
+ 32.5283525085451,
809
+ 34.5494499206543,
810
+ 20.96703338623047,
811
+ 18.90109634399414
812
+ ],
813
+ [
814
+ 27.47252655029297,
815
+ 33.934945831299004,
816
+ 36.21977996826172,
817
+ 21.4945068359375,
818
+ 19.428573608398438
819
+ ]
820
+ ]
821
+ }
822
+ }
823
+ }
824
+ }
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": false,
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": 0.2,
53
+ "exclude_state": false,
54
+ "use_mean_std": false,
55
+ "max_num_embodiments": 32
56
+ }
experiment_cfg/final_processor_config.json ADDED
The diff for this file is too large to render. See raw diff
 
model-00001-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cbcaea5ee88f1e0f1465043920a2647c67e7de17d24adfd1c477742a6168edec
3
+ size 4986649584
model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9790ca026ec64a381ea6c05602232467bd45678e243417e645c0474cecf146e4
3
+ size 4970792616
model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:09a25bd86f40e2e38be2efeeeabd38b95ee58cca3cde5db7107c026747129d2e
3
+ size 2618758696
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
processor_config.json ADDED
@@ -0,0 +1,1123 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "front",
1019
+ "wrist"
1020
+ ],
1021
+ "sin_cos_embedding_keys": null,
1022
+ "mean_std_embedding_keys": null,
1023
+ "action_configs": null
1024
+ },
1025
+ "state": {
1026
+ "delta_indices": [
1027
+ 0
1028
+ ],
1029
+ "modality_keys": [
1030
+ "single_arm",
1031
+ "gripper"
1032
+ ],
1033
+ "sin_cos_embedding_keys": null,
1034
+ "mean_std_embedding_keys": null,
1035
+ "action_configs": null
1036
+ },
1037
+ "action": {
1038
+ "delta_indices": [
1039
+ 0,
1040
+ 1,
1041
+ 2,
1042
+ 3,
1043
+ 4,
1044
+ 5,
1045
+ 6,
1046
+ 7,
1047
+ 8,
1048
+ 9,
1049
+ 10,
1050
+ 11,
1051
+ 12,
1052
+ 13,
1053
+ 14,
1054
+ 15
1055
+ ],
1056
+ "modality_keys": [
1057
+ "single_arm",
1058
+ "gripper"
1059
+ ],
1060
+ "sin_cos_embedding_keys": null,
1061
+ "mean_std_embedding_keys": null,
1062
+ "action_configs": [
1063
+ {
1064
+ "rep": "RELATIVE",
1065
+ "type": "NON_EEF",
1066
+ "format": "DEFAULT",
1067
+ "state_key": null
1068
+ },
1069
+ {
1070
+ "rep": "ABSOLUTE",
1071
+ "type": "NON_EEF",
1072
+ "format": "DEFAULT",
1073
+ "state_key": null
1074
+ }
1075
+ ]
1076
+ },
1077
+ "language": {
1078
+ "delta_indices": [
1079
+ 0
1080
+ ],
1081
+ "modality_keys": [
1082
+ "annotation.human.task_description"
1083
+ ],
1084
+ "sin_cos_embedding_keys": null,
1085
+ "mean_std_embedding_keys": null,
1086
+ "action_configs": null
1087
+ }
1088
+ }
1089
+ },
1090
+ "image_crop_size": [
1091
+ 230,
1092
+ 230
1093
+ ],
1094
+ "image_target_size": [
1095
+ 256,
1096
+ 256
1097
+ ],
1098
+ "use_albumentations": true,
1099
+ "random_rotation_angle": 0,
1100
+ "color_jitter_params": {
1101
+ "brightness": 0.3,
1102
+ "contrast": 0.4,
1103
+ "saturation": 0.5,
1104
+ "hue": 0.08
1105
+ },
1106
+ "shortest_image_edge": 256,
1107
+ "crop_fraction": 0.95,
1108
+ "letter_box_transform": false,
1109
+ "model_name": "nvidia/Cosmos-Reason2-2B",
1110
+ "model_type": "qwen",
1111
+ "formalize_language": true,
1112
+ "max_state_dim": 132,
1113
+ "max_action_dim": 132,
1114
+ "max_action_horizon": 40,
1115
+ "use_percentiles": true,
1116
+ "use_mean_std": false,
1117
+ "clip_outliers": true,
1118
+ "apply_sincos_state_encoding": false,
1119
+ "use_relative_action": true,
1120
+ "exclude_state": false,
1121
+ "state_dropout_prob": 0.2
1122
+ }
1123
+ }
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29a68714263fe335a30564d7d912ed40602e71e8c063eb2fe556a2c688637044
3
+ size 14645
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4db75d62db34e58c117f52804bbcf73f17f0787bfd8c708a85e80b7b061df5df
3
+ size 1465
statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3598e3f183c48504c29dcdfd077980a5ac714fef1fe108d1244511a4f0a86331
3
+ size 5841
wandb_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"project": "finetune-gr00t-n1d7", "run_id": "groot_objects2"}