Hugging Face's logo

pravsels
/
groot1.7_tool_removal

Robotics
Safetensors
gr00t
Gr00tN1d7
gr00t-n1.7
vla
Model card Files
xet
 Community
pravsels commited on
Commit
fb9979d
·
verified ·
1 Parent(s): 18fe184

Upload GR00T N1.7 checkpoint step 6000

Browse files
Files changed (14) hide show
  1. checkpoints/6000/pretrained_model/config.json +88 -0
  2. checkpoints/6000/pretrained_model/embodiment_id.json +57 -0
  3. checkpoints/6000/pretrained_model/experiment_cfg/conf.yaml +219 -0
  4. checkpoints/6000/pretrained_model/experiment_cfg/config.yaml +217 -0
  5. checkpoints/6000/pretrained_model/experiment_cfg/dataset_statistics.json +824 -0
  6. checkpoints/6000/pretrained_model/experiment_cfg/final_model_config.json +56 -0
  7. checkpoints/6000/pretrained_model/experiment_cfg/final_processor_config.json +0 -0
  8. checkpoints/6000/pretrained_model/model-00001-of-00003.safetensors +3 -0
  9. checkpoints/6000/pretrained_model/model-00002-of-00003.safetensors +3 -0
  10. checkpoints/6000/pretrained_model/model-00003-of-00003.safetensors +3 -0
  11. checkpoints/6000/pretrained_model/model.safetensors.index.json +0 -0
  12. checkpoints/6000/pretrained_model/processor_config.json +1124 -0
  13. checkpoints/6000/pretrained_model/statistics.json +0 -0
  14. checkpoints/6000/pretrained_model/wandb_config.json +1 -0
checkpoints/6000/pretrained_model/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
+ }
checkpoints/6000/pretrained_model/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
+ "robocasa_gr1_tabletop": 10,
55
+ "new_embodiment": 10,
56
+ "unitree_g1_sonic": 11
57
+ }
checkpoints/6000/pretrained_model/experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,219 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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:
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: 0.2
69
+ exclude_state: false
70
+ use_mean_std: false
71
+ max_num_embodiments: 32
72
+ data:
73
+ datasets:
74
+ - dataset_paths:
75
+ - /scratch/u6kr/pravsels.u6kr/groot/datasets/tool_removal
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
+ - front
87
+ - wrist
88
+ - top
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
+ - single_arm
97
+ - gripper
98
+ sin_cos_embedding_keys: null
99
+ mean_std_embedding_keys: null
100
+ action_configs: null
101
+ action:
102
+ delta_indices:
103
+ - 0
104
+ - 1
105
+ - 2
106
+ - 3
107
+ - 4
108
+ - 5
109
+ - 6
110
+ - 7
111
+ - 8
112
+ - 9
113
+ - 10
114
+ - 11
115
+ - 12
116
+ - 13
117
+ - 14
118
+ - 15
119
+ modality_keys:
120
+ - single_arm
121
+ - gripper
122
+ sin_cos_embedding_keys: null
123
+ mean_std_embedding_keys: null
124
+ action_configs:
125
+ - rep: RELATIVE
126
+ type: NON_EEF
127
+ format: DEFAULT
128
+ state_key: null
129
+ - rep: ABSOLUTE
130
+ type: NON_EEF
131
+ format: DEFAULT
132
+ state_key: null
133
+ language:
134
+ delta_indices:
135
+ - 0
136
+ modality_keys:
137
+ - annotation.human.task_description
138
+ sin_cos_embedding_keys: null
139
+ mean_std_embedding_keys: null
140
+ action_configs: null
141
+ download_cache: false
142
+ shard_size: 1024
143
+ episode_sampling_rate: 0.1
144
+ num_shards_per_epoch: 100000
145
+ override_pretraining_statistics: true
146
+ mode: single_turn
147
+ random_chop: 0.0
148
+ mock_dataset_mode: false
149
+ shuffle: true
150
+ seed: 42
151
+ multiprocessing_context: fork
152
+ allow_padding: false
153
+ subsample_ratio: 1.0
154
+ image_crop_size:
155
+ - 244
156
+ - 244
157
+ image_target_size:
158
+ - 224
159
+ - 224
160
+ training:
161
+ output_dir: /scratch/u6kr/pravsels.u6kr/groot/checkpoints/groot_tool_removal
162
+ experiment_name: groot_tool_removal
163
+ max_steps: 30000
164
+ global_batch_size: 64
165
+ per_gpu_batch_size: null
166
+ gradient_accumulation_steps: 1
167
+ learning_rate: 0.0001
168
+ lr_scheduler_type: cosine
169
+ weight_decay: 1.0e-05
170
+ warmup_ratio: 0.05
171
+ warmup_steps: 0
172
+ max_grad_norm: 1.0
173
+ optim: adamw_torch
174
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
175
+ skip_weight_loading: false
176
+ tf32: true
177
+ fp16: false
178
+ bf16: true
179
+ eval_bf16: true
180
+ logging_steps: 10
181
+ save_steps: 2000
182
+ save_total_limit: 5
183
+ save_vl_model: false
184
+ save_only_model: false
185
+ resume_from_checkpoint: false
186
+ upload_checkpoints: false
187
+ upload_every: 1000
188
+ upload_last_n_checkpoints: 5
189
+ max_concurrent_uploads: 2
190
+ eval_strategy: 'no'
191
+ eval_steps: 500
192
+ eval_set_split_ratio: 0.1
193
+ eval_batch_size: 2
194
+ save_best_eval_metric_name: ''
195
+ save_best_eval_metric_greater_is_better: true
196
+ deepspeed_stage: 2
197
+ gradient_checkpointing: false
198
+ transformers_trust_remote_code: true
199
+ transformers_local_files_only: false
200
+ transformers_cache_dir: null
201
+ transformers_access_token: null
202
+ use_ddp: false
203
+ ddp_bucket_cap_mb: 100
204
+ num_gpus: 1
205
+ dataloader_num_workers: 8
206
+ remove_unused_columns: false
207
+ use_wandb: true
208
+ wandb_project: groot_tool_removal
209
+ enable_profiling: false
210
+ max_retries: 3
211
+ assert_loss_less_than: null
212
+ add_rl_callback: false
213
+ enable_open_loop_eval: false
214
+ open_loop_eval_traj_ids:
215
+ - 0
216
+ open_loop_eval_steps_per_traj: 100
217
+ open_loop_eval_plot_indices: null
218
+ max_steps: 30000
219
+ save_steps: 2000
checkpoints/6000/pretrained_model/experiment_cfg/config.yaml ADDED
@@ -0,0 +1,217 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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:
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: 0.2
69
+ exclude_state: false
70
+ use_mean_std: false
71
+ max_num_embodiments: 32
72
+ data:
73
+ datasets:
74
+ - dataset_paths:
75
+ - /scratch/u6kr/pravsels.u6kr/groot/datasets/tool_removal
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
+ - front
87
+ - wrist
88
+ - top
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
+ - single_arm
97
+ - gripper
98
+ sin_cos_embedding_keys: null
99
+ mean_std_embedding_keys: null
100
+ action_configs: null
101
+ action:
102
+ delta_indices:
103
+ - 0
104
+ - 1
105
+ - 2
106
+ - 3
107
+ - 4
108
+ - 5
109
+ - 6
110
+ - 7
111
+ - 8
112
+ - 9
113
+ - 10
114
+ - 11
115
+ - 12
116
+ - 13
117
+ - 14
118
+ - 15
119
+ modality_keys:
120
+ - single_arm
121
+ - gripper
122
+ sin_cos_embedding_keys: null
123
+ mean_std_embedding_keys: null
124
+ action_configs:
125
+ - rep: relative
126
+ type: non_eef
127
+ format: default
128
+ state_key: null
129
+ - rep: absolute
130
+ type: non_eef
131
+ format: default
132
+ state_key: null
133
+ language:
134
+ delta_indices:
135
+ - 0
136
+ modality_keys:
137
+ - annotation.human.task_description
138
+ sin_cos_embedding_keys: null
139
+ mean_std_embedding_keys: null
140
+ action_configs: null
141
+ download_cache: false
142
+ shard_size: 1024
143
+ episode_sampling_rate: 0.1
144
+ num_shards_per_epoch: 100000
145
+ override_pretraining_statistics: true
146
+ mode: single_turn
147
+ random_chop: 0.0
148
+ mock_dataset_mode: false
149
+ shuffle: true
150
+ seed: 42
151
+ multiprocessing_context: fork
152
+ allow_padding: false
153
+ subsample_ratio: 1.0
154
+ image_crop_size:
155
+ - 244
156
+ - 244
157
+ image_target_size:
158
+ - 224
159
+ - 224
160
+ training:
161
+ output_dir: /scratch/u6kr/pravsels.u6kr/groot/checkpoints/groot_tool_removal
162
+ experiment_name: groot_tool_removal
163
+ max_steps: 30000
164
+ global_batch_size: 64
165
+ per_gpu_batch_size: null
166
+ gradient_accumulation_steps: 1
167
+ learning_rate: 0.0001
168
+ lr_scheduler_type: cosine
169
+ weight_decay: 1.0e-05
170
+ warmup_ratio: 0.05
171
+ warmup_steps: 0
172
+ max_grad_norm: 1.0
173
+ optim: adamw_torch
174
+ start_from_checkpoint: nvidia/GR00T-N1.7-3B
175
+ skip_weight_loading: false
176
+ tf32: true
177
+ fp16: false
178
+ bf16: true
179
+ eval_bf16: true
180
+ logging_steps: 10
181
+ save_steps: 2000
182
+ save_total_limit: 5
183
+ save_vl_model: false
184
+ save_only_model: false
185
+ resume_from_checkpoint: false
186
+ upload_checkpoints: false
187
+ upload_every: 1000
188
+ upload_last_n_checkpoints: 5
189
+ max_concurrent_uploads: 2
190
+ eval_strategy: 'no'
191
+ eval_steps: 500
192
+ eval_set_split_ratio: 0.1
193
+ eval_batch_size: 2
194
+ save_best_eval_metric_name: ''
195
+ save_best_eval_metric_greater_is_better: true
196
+ deepspeed_stage: 2
197
+ gradient_checkpointing: false
198
+ transformers_trust_remote_code: true
199
+ transformers_local_files_only: false
200
+ transformers_cache_dir: null
201
+ transformers_access_token: null
202
+ use_ddp: false
203
+ ddp_bucket_cap_mb: 100
204
+ num_gpus: 1
205
+ dataloader_num_workers: 8
206
+ remove_unused_columns: false
207
+ use_wandb: true
208
+ wandb_project: groot_tool_removal
209
+ enable_profiling: false
210
+ max_retries: 3
211
+ assert_loss_less_than: null
212
+ add_rl_callback: false
213
+ enable_open_loop_eval: false
214
+ open_loop_eval_traj_ids:
215
+ - 0
216
+ open_loop_eval_steps_per_traj: 100
217
+ open_loop_eval_plot_indices: null
checkpoints/6000/pretrained_model/experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,824 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "single_arm": {
5
+ "min": [
6
+ -35.60439682006836,
7
+ -102.28571319580078,
8
+ -42.5494499206543,
9
+ -46.85714340209961,
10
+ -17.80219841003418
11
+ ],
12
+ "max": [
13
+ 39.912086486816406,
14
+ 53.31867980957031,
15
+ 95.9120864868164,
16
+ 99.6043930053711,
17
+ 111.16483306884766
18
+ ],
19
+ "mean": [
20
+ -9.324679374694824,
21
+ -11.791141510009766,
22
+ 27.482675552368164,
23
+ 61.9761848449707,
24
+ 51.434207916259766
25
+ ],
26
+ "std": [
27
+ 10.595321655273438,
28
+ 45.26567459106445,
29
+ 36.83135986328125,
30
+ 19.38697052001953,
31
+ 34.53913497924805
32
+ ],
33
+ "q01": [
34
+ -28.043956756591797,
35
+ -101.75823974609375,
36
+ -23.360879364013677,
37
+ -39.284395446777346,
38
+ -8.747252464294434
39
+ ],
40
+ "q99": [
41
+ 30.631208667755153,
42
+ 36.829011344909695,
43
+ 95.82417297363281,
44
+ 90.81318664550781,
45
+ 97.45054626464844
46
+ ]
47
+ },
48
+ "gripper": {
49
+ "min": [
50
+ 0.6821282505989075
51
+ ],
52
+ "max": [
53
+ 28.51296043395996
54
+ ],
55
+ "mean": [
56
+ 6.521620273590088
57
+ ],
58
+ "std": [
59
+ 5.619235038757324
60
+ ],
61
+ "q01": [
62
+ 0.9549795389175415
63
+ ],
64
+ "q99": [
65
+ 21.82810401916504
66
+ ]
67
+ }
68
+ },
69
+ "action": {
70
+ "single_arm": {
71
+ "min": [
72
+ -36.0,
73
+ -103.25274658203125,
74
+ -43.340660095214844,
75
+ -47.78022003173828,
76
+ -19.91208839416504
77
+ ],
78
+ "max": [
79
+ 40.483516693115234,
80
+ 53.58241653442383,
81
+ 97.31867980957031,
82
+ 100.35164642333984,
83
+ 112.21977996826172
84
+ ],
85
+ "mean": [
86
+ -9.366114616394043,
87
+ -13.918828964233398,
88
+ 26.146936416625977,
89
+ 61.763282775878906,
90
+ 51.52940368652344
91
+ ],
92
+ "std": [
93
+ 10.772896766662598,
94
+ 44.42418670654297,
95
+ 37.867000579833984,
96
+ 19.50536346435547,
97
+ 34.643184661865234
98
+ ],
99
+ "q01": [
100
+ -28.0,
101
+ -102.46154022216797,
102
+ -25.60791172027588,
103
+ -39.07692337036133,
104
+ -9.048791179656982
105
+ ],
106
+ "q99": [
107
+ 31.076923370361328,
108
+ 35.29670333862305,
109
+ 97.23076629638672,
110
+ 91.0707694244385,
111
+ 99.20879364013672
112
+ ]
113
+ },
114
+ "gripper": {
115
+ "min": [
116
+ 0.0
117
+ ],
118
+ "max": [
119
+ 29.106857299804688
120
+ ],
121
+ "mean": [
122
+ 5.589422702789307
123
+ ],
124
+ "std": [
125
+ 6.2185258865356445
126
+ ],
127
+ "q01": [
128
+ 0.6379585266113281
129
+ ],
130
+ "q99": [
131
+ 22.56778335571289
132
+ ]
133
+ }
134
+ },
135
+ "relative_action": {
136
+ "single_arm": {
137
+ "min": [
138
+ [
139
+ -14.725275039672852,
140
+ -39.736263275146484,
141
+ -34.5494499206543,
142
+ -17.274723052978516,
143
+ -50.81318664550781
144
+ ],
145
+ [
146
+ -19.4725284576416,
147
+ -45.53845977783203,
148
+ -38.41758346557617,
149
+ -22.197803497314453,
150
+ -60.92308044433594
151
+ ],
152
+ [
153
+ -22.285715103149414,
154
+ -50.46154022216797,
155
+ -42.4615364074707,
156
+ -26.85714340209961,
157
+ -69.80220031738281
158
+ ],
159
+ [
160
+ -26.065935134887695,
161
+ -55.20879364013672,
162
+ -46.85714340209961,
163
+ -30.901100158691406,
164
+ -78.32967376708984
165
+ ],
166
+ [
167
+ -30.0219783782959,
168
+ -60.0439567565918,
169
+ -50.90109634399414,
170
+ -34.68132019042969,
171
+ -86.5934066772461
172
+ ],
173
+ [
174
+ -32.57143020629883,
175
+ -64.0,
176
+ -54.06593704223633,
177
+ -38.28571319580078,
178
+ -92.83515930175781
179
+ ],
180
+ [
181
+ -34.94505310058594,
182
+ -68.13186645507812,
183
+ -57.31867980957031,
184
+ -41.626373291015625,
185
+ -98.46153259277344
186
+ ],
187
+ [
188
+ -37.4945068359375,
189
+ -71.64835357666016,
190
+ -61.362632751464844,
191
+ -44.70329284667969,
192
+ -101.71428680419922
193
+ ],
194
+ [
195
+ -39.252750396728516,
196
+ -75.6043930053711,
197
+ -64.52747344970703,
198
+ -47.42856979370117,
199
+ -104.8791275024414
200
+ ],
201
+ [
202
+ -40.74725341796875,
203
+ -79.56044006347656,
204
+ -69.36264038085938,
205
+ -49.62636947631836,
206
+ -106.63736724853516
207
+ ],
208
+ [
209
+ -41.978023529052734,
210
+ -83.25274658203125,
211
+ -73.4065933227539,
212
+ -52.52747344970703,
213
+ -109.45054626464844
214
+ ],
215
+ [
216
+ -43.47252655029297,
217
+ -86.76923370361328,
218
+ -77.62637329101562,
219
+ -55.86812973022461,
220
+ -111.73625946044922
221
+ ],
222
+ [
223
+ -44.263736724853516,
224
+ -90.02198028564453,
225
+ -80.17582702636719,
226
+ -58.68131637573242,
227
+ -113.31867980957031
228
+ ],
229
+ [
230
+ -44.96703338623047,
231
+ -92.57142639160156,
232
+ -82.28571319580078,
233
+ -60.96703338623047,
234
+ -113.58241271972656
235
+ ],
236
+ [
237
+ -45.93406295776367,
238
+ -95.20879364013672,
239
+ -84.74725341796875,
240
+ -62.989009857177734,
241
+ -113.75823974609375
242
+ ],
243
+ [
244
+ -45.93406295776367,
245
+ -97.93406677246094,
246
+ -87.29670715332031,
247
+ -64.65934753417969,
248
+ -113.75823974609375
249
+ ]
250
+ ],
251
+ "max": [
252
+ [
253
+ 21.582416534423828,
254
+ 38.06593704223633,
255
+ 29.714282989501953,
256
+ 20.791210174560547,
257
+ 61.45054626464844
258
+ ],
259
+ [
260
+ 25.626373291015625,
261
+ 43.42856979370117,
262
+ 35.164833068847656,
263
+ 27.384614944458008,
264
+ 71.12088012695312
265
+ ],
266
+ [
267
+ 29.054943084716797,
268
+ 48.000003814697266,
269
+ 40.26373291015625,
270
+ 33.97801971435547,
271
+ 80.79120635986328
272
+ ],
273
+ [
274
+ 31.868131637573242,
275
+ 52.13187026977539,
276
+ 44.923072814941406,
277
+ 40.483516693115234,
278
+ 88.5274658203125
279
+ ],
280
+ [
281
+ 36.17582321166992,
282
+ 56.000003814697266,
283
+ 49.406593322753906,
284
+ 46.901100158691406,
285
+ 94.32966613769531
286
+ ],
287
+ [
288
+ 40.21977996826172,
289
+ 59.252750396728516,
290
+ 53.626373291015625,
291
+ 52.879119873046875,
292
+ 98.19779968261719
293
+ ],
294
+ [
295
+ 43.64834976196289,
296
+ 63.912086486816406,
297
+ 57.05494689941406,
298
+ 58.24176025390625,
299
+ 100.13186645507812
300
+ ],
301
+ [
302
+ 46.28571319580078,
303
+ 68.83516693115234,
304
+ 60.659339904785156,
305
+ 62.901100158691406,
306
+ 102.15384674072266
307
+ ],
308
+ [
309
+ 49.4505500793457,
310
+ 74.72527313232422,
311
+ 63.47252655029297,
312
+ 67.03296661376953,
313
+ 105.14286041259766
314
+ ],
315
+ [
316
+ 51.91209030151367,
317
+ 79.3846206665039,
318
+ 66.19779968261719,
319
+ 70.989013671875,
320
+ 106.72527313232422
321
+ ],
322
+ [
323
+ 54.46154022216797,
324
+ 81.23077392578125,
325
+ 69.010986328125,
326
+ 74.32966613769531,
327
+ 107.95603942871094
328
+ ],
329
+ [
330
+ 56.74725341796875,
331
+ 84.30769348144531,
332
+ 71.82417297363281,
333
+ 77.4065933227539,
334
+ 108.21977996826172
335
+ ],
336
+ [
337
+ 59.20879364013672,
338
+ 85.80220031738281,
339
+ 73.84615325927734,
340
+ 79.42857360839844,
341
+ 108.21977996826172
342
+ ],
343
+ [
344
+ 61.230770111083984,
345
+ 88.0,
346
+ 76.3956069946289,
347
+ 81.62637329101562,
348
+ 108.21977996826172
349
+ ],
350
+ [
351
+ 62.81318664550781,
352
+ 90.63735961914062,
353
+ 78.41758728027344,
354
+ 83.73626708984375,
355
+ 108.30769348144531
356
+ ],
357
+ [
358
+ 64.04396057128906,
359
+ 92.92308044433594,
360
+ 80.43955993652344,
361
+ 86.28571319580078,
362
+ 108.3956069946289
363
+ ]
364
+ ],
365
+ "mean": [
366
+ [
367
+ -0.049158670008182526,
368
+ -2.168941020965576,
369
+ -1.4913849830627441,
370
+ -0.3559165298938751,
371
+ 0.09202209115028381
372
+ ],
373
+ [
374
+ -0.05084080994129181,
375
+ -2.165224552154541,
376
+ -1.495241641998291,
377
+ -0.39331620931625366,
378
+ 0.09121254831552505
379
+ ],
380
+ [
381
+ -0.05207787826657295,
382
+ -2.163196086883545,
383
+ -1.4974784851074219,
384
+ -0.4263301491737366,
385
+ 0.09072688221931458
386
+ ],
387
+ [
388
+ -0.05296807736158371,
389
+ -2.162205219268799,
390
+ -1.498586654663086,
391
+ -0.45452311635017395,
392
+ 0.09050723910331726
393
+ ],
394
+ [
395
+ -0.053586602210998535,
396
+ -2.1617705821990967,
397
+ -1.4990029335021973,
398
+ -0.4773682653903961,
399
+ 0.09048400819301605
400
+ ],
401
+ [
402
+ -0.05414744094014168,
403
+ -2.1614644527435303,
404
+ -1.4991257190704346,
405
+ -0.4943642318248749,
406
+ 0.09057102352380753
407
+ ],
408
+ [
409
+ -0.05453475937247276,
410
+ -2.1612329483032227,
411
+ -1.4989585876464844,
412
+ -0.5053240060806274,
413
+ 0.09084251523017883
414
+ ],
415
+ [
416
+ -0.054852601140737534,
417
+ -2.161111831665039,
418
+ -1.4984205961227417,
419
+ -0.510614812374115,
420
+ 0.09114309400320053
421
+ ],
422
+ [
423
+ -0.0554075613617897,
424
+ -2.1620259284973145,
425
+ -1.4965649843215942,
426
+ -0.5129682421684265,
427
+ 0.09149609506130219
428
+ ],
429
+ [
430
+ -0.05622830241918564,
431
+ -2.1631851196289062,
432
+ -1.4943227767944336,
433
+ -0.5145121812820435,
434
+ 0.09184877574443817
435
+ ],
436
+ [
437
+ -0.05717643350362778,
438
+ -2.1646478176116943,
439
+ -1.4914681911468506,
440
+ -0.5157256126403809,
441
+ 0.09221862256526947
442
+ ],
443
+ [
444
+ -0.058130279183387756,
445
+ -2.1664230823516846,
446
+ -1.4880328178405762,
447
+ -0.5167609453201294,
448
+ 0.09259450435638428
449
+ ],
450
+ [
451
+ -0.05908399820327759,
452
+ -2.168487071990967,
453
+ -1.4839659929275513,
454
+ -0.5176403522491455,
455
+ 0.09297100454568863
456
+ ],
457
+ [
458
+ -0.05999736860394478,
459
+ -2.1707944869995117,
460
+ -1.4793920516967773,
461
+ -0.5183396339416504,
462
+ 0.0933464840054512
463
+ ],
464
+ [
465
+ -0.060904815793037415,
466
+ -2.1735663414001465,
467
+ -1.474161148071289,
468
+ -0.5188484191894531,
469
+ 0.09372241795063019
470
+ ],
471
+ [
472
+ -0.06178923323750496,
473
+ -2.1768906116485596,
474
+ -1.4683825969696045,
475
+ -0.5192939639091492,
476
+ 0.09409775584936142
477
+ ]
478
+ ],
479
+ "std": [
480
+ [
481
+ 2.4846138954162598,
482
+ 6.08482551574707,
483
+ 4.9501633644104,
484
+ 2.6457762718200684,
485
+ 7.580384254455566
486
+ ],
487
+ [
488
+ 3.170689821243286,
489
+ 7.566461086273193,
490
+ 6.293671607971191,
491
+ 3.4073100090026855,
492
+ 9.497262954711914
493
+ ],
494
+ [
495
+ 3.844189405441284,
496
+ 9.065008163452148,
497
+ 7.630911350250244,
498
+ 4.137608528137207,
499
+ 11.326314926147461
500
+ ],
501
+ [
502
+ 4.497386932373047,
503
+ 10.561771392822266,
504
+ 8.951238632202148,
505
+ 4.835171222686768,
506
+ 13.058839797973633
507
+ ],
508
+ [
509
+ 5.126635551452637,
510
+ 12.046217918395996,
511
+ 10.248961448669434,
512
+ 5.502501487731934,
513
+ 14.692830085754395
514
+ ],
515
+ [
516
+ 5.729096412658691,
517
+ 13.512539863586426,
518
+ 11.520867347717285,
519
+ 6.143070220947266,
520
+ 16.229785919189453
521
+ ],
522
+ [
523
+ 6.304162979125977,
524
+ 14.957094192504883,
525
+ 12.765586853027344,
526
+ 6.759618282318115,
527
+ 17.674062728881836
528
+ ],
529
+ [
530
+ 6.851644515991211,
531
+ 16.378002166748047,
532
+ 13.982623100280762,
533
+ 7.352746486663818,
534
+ 19.031139373779297
535
+ ],
536
+ [
537
+ 7.372150421142578,
538
+ 17.773666381835938,
539
+ 15.170845031738281,
540
+ 7.916732311248779,
541
+ 20.3072566986084
542
+ ],
543
+ [
544
+ 7.866299152374268,
545
+ 19.143878936767578,
546
+ 16.331148147583008,
547
+ 8.44809341430664,
548
+ 21.508644104003906
549
+ ],
550
+ [
551
+ 8.334341049194336,
552
+ 20.488201141357422,
553
+ 17.46393394470215,
554
+ 8.94785213470459,
555
+ 22.64142608642578
556
+ ],
557
+ [
558
+ 8.776926040649414,
559
+ 21.80654525756836,
560
+ 18.569456100463867,
561
+ 9.418201446533203,
562
+ 23.711524963378906
563
+ ],
564
+ [
565
+ 9.19458293914795,
566
+ 23.099279403686523,
567
+ 19.64833641052246,
568
+ 9.861288070678711,
569
+ 24.72452163696289
570
+ ],
571
+ [
572
+ 9.587946891784668,
573
+ 24.36676025390625,
574
+ 20.701452255249023,
575
+ 10.279257774353027,
576
+ 25.685462951660156
577
+ ],
578
+ [
579
+ 9.957334518432617,
580
+ 25.609230041503906,
581
+ 21.729270935058594,
582
+ 10.674188613891602,
583
+ 26.599021911621094
584
+ ],
585
+ [
586
+ 10.303248405456543,
587
+ 26.827695846557617,
588
+ 22.733097076416016,
589
+ 11.047439575195312,
590
+ 27.469398498535156
591
+ ]
592
+ ],
593
+ "q01": [
594
+ [
595
+ -8.483516693115234,
596
+ -15.912090301513672,
597
+ -16.872966194152834,
598
+ -9.538463592529297,
599
+ -25.846153259277344
600
+ ],
601
+ [
602
+ -11.114725427627564,
603
+ -19.95604705810547,
604
+ -21.53846155166626,
605
+ -12.609229469299317,
606
+ -32.79120990753174
607
+ ],
608
+ [
609
+ -13.664176168441774,
610
+ -24.17582498550415,
611
+ -26.197799682617188,
612
+ -15.24658962249756,
613
+ -40.081756019592284
614
+ ],
615
+ [
616
+ -16.131868362426758,
617
+ -28.21977996826172,
618
+ -30.59340654373169,
619
+ -17.796042823791506,
620
+ -45.883956336975096
621
+ ],
622
+ [
623
+ -18.593406677246094,
624
+ -32.351646423339844,
625
+ -35.06461574554444,
626
+ -20.175822944641112,
627
+ -51.95604297637939
628
+ ],
629
+ [
630
+ -20.52747344970703,
631
+ -36.4835205078125,
632
+ -39.20879364013672,
633
+ -22.285717010498047,
634
+ -57.03648532867432
635
+ ],
636
+ [
637
+ -22.71296564102173,
638
+ -40.263736457824706,
639
+ -43.51648685455322,
640
+ -24.043956756591797,
641
+ -62.24175998687744
642
+ ],
643
+ [
644
+ -24.213626155853273,
645
+ -43.95604679107666,
646
+ -47.642196083068846,
647
+ -26.417583198547362,
648
+ -66.62505226135254
649
+ ],
650
+ [
651
+ -26.14153720855713,
652
+ -47.64219989776611,
653
+ -51.510325813293456,
654
+ -28.075605754852297,
655
+ -70.05362533569335
656
+ ],
657
+ [
658
+ -27.378461132049562,
659
+ -51.3345064163208,
660
+ -54.94505664825439,
661
+ -29.048789405822756,
662
+ -74.19164627075196
663
+ ],
664
+ [
665
+ -28.703296661376953,
666
+ -54.50549289703369,
667
+ -58.45538272857666,
668
+ -30.53098834991455,
669
+ -77.53845977783203
670
+ ],
671
+ [
672
+ -29.66417610168457,
673
+ -57.67032978057861,
674
+ -61.45054981231689,
675
+ -31.60439655303955,
676
+ -80.52747291564941
677
+ ],
678
+ [
679
+ -30.631207714080812,
680
+ -60.91691940307617,
681
+ -64.43955993652344,
682
+ -32.57142639160156,
683
+ -83.24659317016602
684
+ ],
685
+ [
686
+ -31.780219898223876,
687
+ -63.64835330963135,
688
+ -67.24659317016602,
689
+ -34.05977954864502,
690
+ -85.18681335449219
691
+ ],
692
+ [
693
+ -32.74725315093994,
694
+ -66.46153968811035,
695
+ -70.22944526672363,
696
+ -35.466372604370115,
697
+ -87.29055320739747
698
+ ],
699
+ [
700
+ -33.62021961212158,
701
+ -69.0989069366455,
702
+ -72.52131950378418,
703
+ -36.860662117004395,
704
+ -88.52131950378418
705
+ ]
706
+ ],
707
+ "q99": [
708
+ [
709
+ 8.835165023803711,
710
+ 15.82417474746704,
711
+ 11.16484043121338,
712
+ 7.164833068847656,
713
+ 35.24043922424322
714
+ ],
715
+ [
716
+ 11.824175834655762,
717
+ 20.923076362609862,
718
+ 15.032966613769531,
719
+ 9.186809539794922,
720
+ 43.32835220336919
721
+ ],
722
+ [
723
+ 14.725274085998535,
724
+ 25.582420082092284,
725
+ 18.813186645507812,
726
+ 11.120880126953125,
727
+ 49.83999931335452
728
+ ],
729
+ [
730
+ 17.614065666198783,
731
+ 30.329669952392578,
732
+ 22.68131814956665,
733
+ 12.967033386230469,
734
+ 56.78505596160891
735
+ ],
736
+ [
737
+ 20.263736724853516,
738
+ 35.076926918029784,
739
+ 26.1916477775574,
740
+ 15.164833068847656,
741
+ 63.120879859924315
742
+ ],
743
+ [
744
+ 22.725274906158447,
745
+ 39.56044361114502,
746
+ 29.620221385955837,
747
+ 16.916919403076196,
748
+ 68.65318649291994
749
+ ],
750
+ [
751
+ 25.098899841308594,
752
+ 43.95604297637939,
753
+ 32.9670295715332,
754
+ 18.85714340209961,
755
+ 73.31867980957031
756
+ ],
757
+ [
758
+ 27.384614944458008,
759
+ 47.9059330749512,
760
+ 36.13187026977539,
761
+ 20.78505622863772,
762
+ 77.27472686767578
763
+ ],
764
+ [
765
+ 29.582418308258056,
766
+ 51.8619763183594,
767
+ 39.03297016143799,
768
+ 22.36747276306155,
769
+ 80.25757949829104
770
+ ],
771
+ [
772
+ 31.516483306884766,
773
+ 55.55428611755374,
774
+ 41.92791282653811,
775
+ 23.95604679107666,
776
+ 83.07691955566406
777
+ ],
778
+ [
779
+ 33.18065967559817,
780
+ 59.07692337036133,
781
+ 44.835166664123534,
782
+ 25.450550079345703,
783
+ 84.92307991027832
784
+ ],
785
+ [
786
+ 35.026812667846706,
787
+ 62.50550025939941,
788
+ 47.736263275146484,
789
+ 26.32966968536377,
790
+ 87.55428611755373
791
+ ],
792
+ [
793
+ 36.25758277893069,
794
+ 65.83999931335451,
795
+ 50.19164955139163,
796
+ 27.296707153320312,
797
+ 89.57010536193853
798
+ ],
799
+ [
800
+ 37.664176368713406,
801
+ 69.18681335449219,
802
+ 53.00483592987063,
803
+ 27.73010959625247,
804
+ 90.72527313232422
805
+ ],
806
+ [
807
+ 38.8070329666138,
808
+ 72.16966598510744,
809
+ 55.37230567932134,
810
+ 28.351646423339844,
811
+ 91.25274658203125
812
+ ],
813
+ [
814
+ 40.389449501037625,
815
+ 75.07076614379885,
816
+ 57.31867980957031,
817
+ 29.048792953491237,
818
+ 92.04395294189453
819
+ ]
820
+ ]
821
+ }
822
+ }
823
+ }
824
+ }
checkpoints/6000/pretrained_model/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
+ }
checkpoints/6000/pretrained_model/experiment_cfg/final_processor_config.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/6000/pretrained_model/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
checkpoints/6000/pretrained_model/model-00002-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92f021dde787d8e8e150532f81ea4c05e9253fd3696821dbe79c556f0b2d29e2
3
+ size 4970792616
checkpoints/6000/pretrained_model/model-00003-of-00003.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:322777cc4b4590f7f88558b55cbb137400416f3f66a93e862c108df085744a40
3
+ size 2618758696
checkpoints/6000/pretrained_model/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/6000/pretrained_model/processor_config.json ADDED
@@ -0,0 +1,1124 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "top"
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
+ "single_arm",
1032
+ "gripper"
1033
+ ],
1034
+ "sin_cos_embedding_keys": null,
1035
+ "mean_std_embedding_keys": null,
1036
+ "action_configs": null
1037
+ },
1038
+ "action": {
1039
+ "delta_indices": [
1040
+ 0,
1041
+ 1,
1042
+ 2,
1043
+ 3,
1044
+ 4,
1045
+ 5,
1046
+ 6,
1047
+ 7,
1048
+ 8,
1049
+ 9,
1050
+ 10,
1051
+ 11,
1052
+ 12,
1053
+ 13,
1054
+ 14,
1055
+ 15
1056
+ ],
1057
+ "modality_keys": [
1058
+ "single_arm",
1059
+ "gripper"
1060
+ ],
1061
+ "sin_cos_embedding_keys": null,
1062
+ "mean_std_embedding_keys": null,
1063
+ "action_configs": [
1064
+ {
1065
+ "rep": "RELATIVE",
1066
+ "type": "NON_EEF",
1067
+ "format": "DEFAULT",
1068
+ "state_key": null
1069
+ },
1070
+ {
1071
+ "rep": "ABSOLUTE",
1072
+ "type": "NON_EEF",
1073
+ "format": "DEFAULT",
1074
+ "state_key": null
1075
+ }
1076
+ ]
1077
+ },
1078
+ "language": {
1079
+ "delta_indices": [
1080
+ 0
1081
+ ],
1082
+ "modality_keys": [
1083
+ "annotation.human.task_description"
1084
+ ],
1085
+ "sin_cos_embedding_keys": null,
1086
+ "mean_std_embedding_keys": null,
1087
+ "action_configs": null
1088
+ }
1089
+ }
1090
+ },
1091
+ "image_crop_size": [
1092
+ 230,
1093
+ 230
1094
+ ],
1095
+ "image_target_size": [
1096
+ 256,
1097
+ 256
1098
+ ],
1099
+ "use_albumentations": true,
1100
+ "random_rotation_angle": 0,
1101
+ "color_jitter_params": {
1102
+ "brightness": 0.3,
1103
+ "contrast": 0.4,
1104
+ "saturation": 0.5,
1105
+ "hue": 0.08
1106
+ },
1107
+ "shortest_image_edge": 256,
1108
+ "crop_fraction": 0.95,
1109
+ "letter_box_transform": false,
1110
+ "model_name": "nvidia/Cosmos-Reason2-2B",
1111
+ "model_type": "qwen",
1112
+ "formalize_language": true,
1113
+ "max_state_dim": 132,
1114
+ "max_action_dim": 132,
1115
+ "max_action_horizon": 40,
1116
+ "use_percentiles": true,
1117
+ "use_mean_std": false,
1118
+ "clip_outliers": true,
1119
+ "apply_sincos_state_encoding": false,
1120
+ "use_relative_action": true,
1121
+ "exclude_state": false,
1122
+ "state_dropout_prob": 0.2
1123
+ }
1124
+ }
checkpoints/6000/pretrained_model/statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoints/6000/pretrained_model/wandb_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"project": "groot_tool_removal", "run_id": "groot_tool_removal"}