Hugging Face's logo

pravsels
/
groot1.7_tool_insert

Robotics
Safetensors
gr00t
Gr00tN1d7
gr00t-n1.7
vla
Model card Files
xet
 Community
pravsels commited on
Commit
e02f7ed
·
verified ·
1 Parent(s): 4b25d74

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_insert
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_insert
162
+ experiment_name: groot_tool_insert
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_insert
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_insert
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_insert
162
+ experiment_name: groot_tool_insert
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_insert
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
+ -63.38461685180664,
7
+ -102.46154022216797,
8
+ -26.461538314819336,
9
+ 7.296703338623047,
10
+ -66.24176025390625
11
+ ],
12
+ "max": [
13
+ 54.241756439208984,
14
+ 37.58241653442383,
15
+ 96.0879135131836,
16
+ 99.6043930053711,
17
+ 155.03297424316406
18
+ ],
19
+ "mean": [
20
+ -9.631865501403809,
21
+ -17.412471771240234,
22
+ 37.66987609863281,
23
+ 61.66801834106445,
24
+ 54.51057815551758
25
+ ],
26
+ "std": [
27
+ 19.43613624572754,
28
+ 39.17250442504883,
29
+ 32.67021942138672,
30
+ 14.489995002746582,
31
+ 33.108238220214844
32
+ ],
33
+ "q01": [
34
+ -50.01758098602295,
35
+ -101.4945068359375,
36
+ -12.654945230484008,
37
+ 18.0219783782959,
38
+ -10.505494117736816
39
+ ],
40
+ "q99": [
41
+ 49.230770111083984,
42
+ 28.879121780395508,
43
+ 96.0,
44
+ 90.10989379882812,
45
+ 108.86593322753926
46
+ ]
47
+ },
48
+ "gripper": {
49
+ "min": [
50
+ 0.6139154434204102
51
+ ],
52
+ "max": [
53
+ 27.899045944213867
54
+ ],
55
+ "mean": [
56
+ 6.210718154907227
57
+ ],
58
+ "std": [
59
+ 5.811882495880127
60
+ ],
61
+ "q01": [
62
+ 0.6139154434204102
63
+ ],
64
+ "q99": [
65
+ 27.899045944213867
66
+ ]
67
+ }
68
+ },
69
+ "action": {
70
+ "single_arm": {
71
+ "min": [
72
+ -63.9560432434082,
73
+ -103.42857360839844,
74
+ -29.626373291015625,
75
+ 5.934065818786621,
76
+ -66.5934066772461
77
+ ],
78
+ "max": [
79
+ 54.72527313232422,
80
+ 36.79121017456055,
81
+ 97.4065933227539,
82
+ 100.0,
83
+ 155.38461303710938
84
+ ],
85
+ "mean": [
86
+ -9.48980712890625,
87
+ -17.977977752685547,
88
+ 36.75624084472656,
89
+ 61.3520622253418,
90
+ 54.54677200317383
91
+ ],
92
+ "std": [
93
+ 19.44814682006836,
94
+ 39.27090835571289,
95
+ 33.18235397338867,
96
+ 14.594162940979004,
97
+ 33.16144561767578
98
+ ],
99
+ "q01": [
100
+ -50.06593322753906,
101
+ -102.28571319580078,
102
+ -14.94505500793457,
103
+ 17.36263656616211,
104
+ -10.681318283081055
105
+ ],
106
+ "q99": [
107
+ 49.18681335449219,
108
+ 28.435164356231752,
109
+ 97.23076629638672,
110
+ 90.15384674072266,
111
+ 109.4945068359375
112
+ ]
113
+ },
114
+ "gripper": {
115
+ "min": [
116
+ 0.23923444747924805
117
+ ],
118
+ "max": [
119
+ 28.3891544342041
120
+ ],
121
+ "mean": [
122
+ 5.042398929595947
123
+ ],
124
+ "std": [
125
+ 6.294958114624023
126
+ ],
127
+ "q01": [
128
+ 0.6379585266113281
129
+ ],
130
+ "q99": [
131
+ 28.070175170898438
132
+ ]
133
+ }
134
+ },
135
+ "relative_action": {
136
+ "single_arm": {
137
+ "min": [
138
+ [
139
+ -16.043954849243164,
140
+ -35.42856979370117,
141
+ -32.79121017456055,
142
+ -19.296707153320312,
143
+ -49.93406677246094
144
+ ],
145
+ [
146
+ -21.14285659790039,
147
+ -39.73625946044922,
148
+ -34.46154022216797,
149
+ -23.78022003173828,
150
+ -55.29670333862305
151
+ ],
152
+ [
153
+ -25.80219841003418,
154
+ -43.69230651855469,
155
+ -39.38461685180664,
156
+ -29.05494499206543,
157
+ -59.604393005371094
158
+ ],
159
+ [
160
+ -29.58241844177246,
161
+ -47.648345947265625,
162
+ -43.69230651855469,
163
+ -34.24175262451172,
164
+ -61.97801971435547
165
+ ],
166
+ [
167
+ -33.4505500793457,
168
+ -51.69230651855469,
169
+ -47.38461685180664,
170
+ -38.637359619140625,
171
+ -64.70330047607422
172
+ ],
173
+ [
174
+ -36.879119873046875,
175
+ -55.648353576660156,
176
+ -50.54945373535156,
177
+ -42.68132019042969,
178
+ -69.62637329101562
179
+ ],
180
+ [
181
+ -39.60439682006836,
182
+ -59.604400634765625,
183
+ -53.5384635925293,
184
+ -46.54945373535156,
185
+ -73.5824203491211
186
+ ],
187
+ [
188
+ -41.978023529052734,
189
+ -63.56044006347656,
190
+ -56.087913513183594,
191
+ -49.890113830566406,
192
+ -75.0769271850586
193
+ ],
194
+ [
195
+ -44.087913513183594,
196
+ -67.16484069824219,
197
+ -58.10988998413086,
198
+ -53.05494689941406,
199
+ -78.5054931640625
200
+ ],
201
+ [
202
+ -46.02198028564453,
203
+ -71.82417297363281,
204
+ -59.60439682006836,
205
+ -55.60439682006836,
206
+ -82.28571319580078
207
+ ],
208
+ [
209
+ -47.78022003173828,
210
+ -76.21977996826172,
211
+ -60.74725341796875,
212
+ -58.0659294128418,
213
+ -85.27471923828125
214
+ ],
215
+ [
216
+ -49.45054626464844,
217
+ -80.26373291015625,
218
+ -61.978023529052734,
219
+ -59.912086486816406,
220
+ -87.03296661376953
221
+ ],
222
+ [
223
+ -50.593406677246094,
224
+ -84.13186645507812,
225
+ -66.02198028564453,
226
+ -61.318687438964844,
227
+ -88.52747344970703
228
+ ],
229
+ [
230
+ -51.82417297363281,
231
+ -88.0,
232
+ -70.32966613769531,
233
+ -62.02198028564453,
234
+ -90.46154022216797
235
+ ],
236
+ [
237
+ -53.58241653442383,
238
+ -91.69230651855469,
239
+ -74.54945373535156,
240
+ -62.373626708984375,
241
+ -91.86813354492188
242
+ ],
243
+ [
244
+ -55.42856979370117,
245
+ -95.03296661376953,
246
+ -78.32966613769531,
247
+ -62.021976470947266,
248
+ -92.57142639160156
249
+ ]
250
+ ],
251
+ "max": [
252
+ [
253
+ 11.648351669311523,
254
+ 24.175823211669922,
255
+ 34.021976470947266,
256
+ 19.03296661376953,
257
+ 56.61538314819336
258
+ ],
259
+ [
260
+ 15.340660095214844,
261
+ 27.956039428710938,
262
+ 37.80219650268555,
263
+ 24.043954849243164,
264
+ 66.54945373535156
265
+ ],
266
+ [
267
+ 19.03296661376953,
268
+ 33.0549430847168,
269
+ 41.31868362426758,
270
+ 29.318681716918945,
271
+ 75.86813354492188
272
+ ],
273
+ [
274
+ 22.10988998413086,
275
+ 38.24176025390625,
276
+ 45.010990142822266,
277
+ 34.153846740722656,
278
+ 83.16483306884766
279
+ ],
280
+ [
281
+ 24.835163116455078,
282
+ 43.34066390991211,
283
+ 51.252750396728516,
284
+ 38.81318664550781,
285
+ 90.37362670898438
286
+ ],
287
+ [
288
+ 27.560440063476562,
289
+ 47.56044387817383,
290
+ 57.40659713745117,
291
+ 43.384613037109375,
292
+ 97.31867980957031
293
+ ],
294
+ [
295
+ 31.516483306884766,
296
+ 51.69231033325195,
297
+ 63.47252655029297,
298
+ 48.30769348144531,
299
+ 101.010986328125
300
+ ],
301
+ [
302
+ 34.41758346557617,
303
+ 54.857147216796875,
304
+ 69.80220031738281,
305
+ 52.96703338623047,
306
+ 103.47252655029297
307
+ ],
308
+ [
309
+ 36.52747344970703,
310
+ 57.846153259277344,
311
+ 75.78022003173828,
312
+ 57.80219650268555,
313
+ 104.87911987304688
314
+ ],
315
+ [
316
+ 39.07692337036133,
317
+ 59.69231033325195,
318
+ 80.61538696289062,
319
+ 61.75823974609375,
320
+ 109.36264038085938
321
+ ],
322
+ [
323
+ 40.483516693115234,
324
+ 62.68131637573242,
325
+ 85.18681335449219,
326
+ 63.516483306884766,
327
+ 114.90110778808594
328
+ ],
329
+ [
330
+ 42.24176025390625,
331
+ 64.61538696289062,
332
+ 89.23076629638672,
333
+ 64.83515930175781,
334
+ 120.35165405273438
335
+ ],
336
+ [
337
+ 44.70329666137695,
338
+ 66.28571319580078,
339
+ 93.010986328125,
340
+ 65.27471923828125,
341
+ 125.89012145996094
342
+ ],
343
+ [
344
+ 46.81318664550781,
345
+ 67.6043930053711,
346
+ 96.08790588378906,
347
+ 65.27471923828125,
348
+ 130.54945373535156
349
+ ],
350
+ [
351
+ 49.71428680419922,
352
+ 69.89010620117188,
353
+ 99.69230651855469,
354
+ 65.0988998413086,
355
+ 135.03297424316406
356
+ ],
357
+ [
358
+ 51.03296661376953,
359
+ 71.64835357666016,
360
+ 102.76922607421875,
361
+ 64.57142639160156,
362
+ 139.5164794921875
363
+ ]
364
+ ],
365
+ "mean": [
366
+ [
367
+ 0.13628681004047394,
368
+ -0.517716646194458,
369
+ -1.0077838897705078,
370
+ -0.5182581543922424,
371
+ 0.03373425453901291
372
+ ],
373
+ [
374
+ 0.13533133268356323,
375
+ -0.5116506814956665,
376
+ -1.0103564262390137,
377
+ -0.5699769854545593,
378
+ 0.03356771543622017
379
+ ],
380
+ [
381
+ 0.1350589394569397,
382
+ -0.5076963901519775,
383
+ -1.011871337890625,
384
+ -0.6158914566040039,
385
+ 0.03350621089339256
386
+ ],
387
+ [
388
+ 0.13525620102882385,
389
+ -0.5053636431694031,
390
+ -1.0124999284744263,
391
+ -0.6553673148155212,
392
+ 0.03365541249513626
393
+ ],
394
+ [
395
+ 0.13557599484920502,
396
+ -0.5040827989578247,
397
+ -1.0125421285629272,
398
+ -0.6877619028091431,
399
+ 0.03388325497508049
400
+ ],
401
+ [
402
+ 0.13577742874622345,
403
+ -0.5033721923828125,
404
+ -1.0125226974487305,
405
+ -0.7124093770980835,
406
+ 0.034220755100250244
407
+ ],
408
+ [
409
+ 0.1358034461736679,
410
+ -0.5028385519981384,
411
+ -1.012499451637268,
412
+ -0.7296768426895142,
413
+ 0.03454918414354324
414
+ ],
415
+ [
416
+ 0.13585159182548523,
417
+ -0.5023515224456787,
418
+ -1.012424349784851,
419
+ -0.7397712469100952,
420
+ 0.03485157713294029
421
+ ],
422
+ [
423
+ 0.13585613667964935,
424
+ -0.5020067095756531,
425
+ -1.01227867603302,
426
+ -0.7451845407485962,
427
+ 0.03513219580054283
428
+ ],
429
+ [
430
+ 0.13573385775089264,
431
+ -0.50176602602005,
432
+ -1.0120726823806763,
433
+ -0.7478693723678589,
434
+ 0.03546963632106781
435
+ ],
436
+ [
437
+ 0.1355406641960144,
438
+ -0.5017039179801941,
439
+ -1.0118080377578735,
440
+ -0.7491047382354736,
441
+ 0.03583425283432007
442
+ ],
443
+ [
444
+ 0.13531775772571564,
445
+ -0.5017746090888977,
446
+ -1.0114587545394897,
447
+ -0.7496314644813538,
448
+ 0.03621070086956024
449
+ ],
450
+ [
451
+ 0.13510282337665558,
452
+ -0.502020537853241,
453
+ -1.010985016822815,
454
+ -0.7496939301490784,
455
+ 0.03658333420753479
456
+ ],
457
+ [
458
+ 0.13488848507404327,
459
+ -0.502433180809021,
460
+ -1.010371208190918,
461
+ -0.7496501207351685,
462
+ 0.0369512215256691
463
+ ],
464
+ [
465
+ 0.13466057181358337,
466
+ -0.5029710531234741,
467
+ -1.0096170902252197,
468
+ -0.7496158480644226,
469
+ 0.03731999918818474
470
+ ],
471
+ [
472
+ 0.13440658152103424,
473
+ -0.5036377310752869,
474
+ -1.0087194442749023,
475
+ -0.749592125415802,
476
+ 0.03768806904554367
477
+ ]
478
+ ],
479
+ "std": [
480
+ [
481
+ 2.3392350673675537,
482
+ 4.905196189880371,
483
+ 4.479983329772949,
484
+ 3.0518853664398193,
485
+ 5.210679054260254
486
+ ],
487
+ [
488
+ 2.9727489948272705,
489
+ 6.244482040405273,
490
+ 5.621309757232666,
491
+ 3.9152672290802,
492
+ 6.68196964263916
493
+ ],
494
+ [
495
+ 3.600023031234741,
496
+ 7.583919048309326,
497
+ 6.75690221786499,
498
+ 4.756592273712158,
499
+ 8.094547271728516
500
+ ],
501
+ [
502
+ 4.214196681976318,
503
+ 8.910733222961426,
504
+ 7.875130653381348,
505
+ 5.5721893310546875,
506
+ 9.444384574890137
507
+ ],
508
+ [
509
+ 4.811962604522705,
510
+ 10.2178373336792,
511
+ 8.969926834106445,
512
+ 6.362659454345703,
513
+ 10.730934143066406
514
+ ],
515
+ [
516
+ 5.39157772064209,
517
+ 11.501067161560059,
518
+ 10.037755966186523,
519
+ 7.129862308502197,
520
+ 11.955413818359375
521
+ ],
522
+ [
523
+ 5.952877521514893,
524
+ 12.758004188537598,
525
+ 11.077422142028809,
526
+ 7.873784065246582,
527
+ 13.119400024414062
528
+ ],
529
+ [
530
+ 6.49616813659668,
531
+ 13.987278938293457,
532
+ 12.088642120361328,
533
+ 8.594879150390625,
534
+ 14.225035667419434
535
+ ],
536
+ [
537
+ 7.021902561187744,
538
+ 15.188458442687988,
539
+ 13.07148265838623,
540
+ 9.283919334411621,
541
+ 15.274985313415527
542
+ ],
543
+ [
544
+ 7.530764102935791,
545
+ 16.36131477355957,
546
+ 14.026312828063965,
547
+ 9.935736656188965,
548
+ 16.272436141967773
549
+ ],
550
+ [
551
+ 8.023550033569336,
552
+ 17.505889892578125,
553
+ 14.953559875488281,
554
+ 10.548323631286621,
555
+ 17.220638275146484
556
+ ],
557
+ [
558
+ 8.501117706298828,
559
+ 18.622634887695312,
560
+ 15.85386848449707,
561
+ 11.121841430664062,
562
+ 18.122875213623047
563
+ ],
564
+ [
565
+ 8.964164733886719,
566
+ 19.711950302124023,
567
+ 16.72788429260254,
568
+ 11.658260345458984,
569
+ 18.982440948486328
570
+ ],
571
+ [
572
+ 9.413512229919434,
573
+ 20.774333953857422,
574
+ 17.576435089111328,
575
+ 12.158870697021484,
576
+ 19.80272102355957
577
+ ],
578
+ [
579
+ 9.850040435791016,
580
+ 21.810409545898438,
581
+ 18.40008544921875,
582
+ 12.625186920166016,
583
+ 20.586706161499023
584
+ ],
585
+ [
586
+ 10.274487495422363,
587
+ 22.82068634033203,
588
+ 19.19965934753418,
589
+ 13.058918952941895,
590
+ 21.33749008178711
591
+ ]
592
+ ],
593
+ "q01": [
594
+ [
595
+ -7.995604944229125,
596
+ -16.263736724853516,
597
+ -13.010988235473633,
598
+ -10.065933227539062,
599
+ -20.30769052505493
600
+ ],
601
+ [
602
+ -10.545056343078612,
603
+ -21.010987186431883,
604
+ -16.351646423339844,
605
+ -13.14285659790039,
606
+ -25.797801208496093
607
+ ],
608
+ [
609
+ -13.142857027053832,
610
+ -25.67032814025879,
611
+ -19.468132972717285,
612
+ -15.956046009063721,
613
+ -31.736263275146484
614
+ ],
615
+ [
616
+ -15.468132972717283,
617
+ -30.241756439208984,
618
+ -22.593406677246094,
619
+ -18.85714340209961,
620
+ -36.57142810821533
621
+ ],
622
+ [
623
+ -17.71428575515747,
624
+ -34.63736515045166,
625
+ -25.494504928588867,
626
+ -21.406595039367676,
627
+ -41.79780292510986
628
+ ],
629
+ [
630
+ -20.039561080932618,
631
+ -38.98461456298828,
632
+ -28.483515644073485,
633
+ -23.907689094543457,
634
+ -45.89011173248291
635
+ ],
636
+ [
637
+ -21.973626136779785,
638
+ -43.340660095214844,
639
+ -31.208789825439453,
640
+ -25.626373291015625,
641
+ -49.846153259277344
642
+ ],
643
+ [
644
+ -24.21977891921997,
645
+ -47.68790912628174,
646
+ -33.75824146270752,
647
+ -27.73626117706299,
648
+ -53.49010944366455
649
+ ],
650
+ [
651
+ -26.329669952392578,
652
+ -51.64395408630371,
653
+ -35.95604496002197,
654
+ -29.230770111083984,
655
+ -56.43955993652344
656
+ ],
657
+ [
658
+ -28.12747287750244,
659
+ -55.424174499511714,
660
+ -38.2813196182251,
661
+ -30.63736047744751,
662
+ -59.60439472198486
663
+ ],
664
+ [
665
+ -29.797801876068114,
666
+ -59.38021945953369,
667
+ -40.351646423339844,
668
+ -31.99560432434082,
669
+ -61.97801971435547
670
+ ],
671
+ [
672
+ -31.516483306884766,
673
+ -62.94505310058594,
674
+ -42.54945373535156,
675
+ -33.27472476959228,
676
+ -64.26373634338378
677
+ ],
678
+ [
679
+ -33.27472686767578,
680
+ -66.5010971069336,
681
+ -44.483516693115234,
682
+ -35.072527694702146,
683
+ -65.92527198791504
684
+ ],
685
+ [
686
+ -34.808791351318355,
687
+ -69.978023147583,
688
+ -46.68131809234619,
689
+ -36.26374053955078,
690
+ -68.21098518371582
691
+ ],
692
+ [
693
+ -36.43955993652344,
694
+ -73.27033157348633,
695
+ -48.7428560256958,
696
+ -37.446152687072754,
697
+ -69.71427917480469
698
+ ],
699
+ [
700
+ -38.06153964996338,
701
+ -76.78681106567383,
702
+ -50.19779968261719,
703
+ -38.501099586486816,
704
+ -70.63296775817871
705
+ ]
706
+ ],
707
+ "q99": [
708
+ [
709
+ 6.8131880283355715,
710
+ 12.043960571289062,
711
+ 14.281319618225162,
712
+ 9.626373291015625,
713
+ 20.659339904785156
714
+ ],
715
+ [
716
+ 8.835165023803711,
717
+ 15.736259460449219,
718
+ 19.204394531250063,
719
+ 12.703296661376953,
720
+ 26.63736343383789
721
+ ],
722
+ [
723
+ 10.94505500793457,
724
+ 19.25274829864502,
725
+ 24.000002765655516,
726
+ 15.780220031738281,
727
+ 32.742856025695865
728
+ ],
729
+ [
730
+ 12.791208696365356,
731
+ 22.5054931640625,
732
+ 28.659339904785156,
733
+ 18.89670171737677,
734
+ 38.32966995239258
735
+ ],
736
+ [
737
+ 14.725275039672852,
738
+ 25.846153259277344,
739
+ 32.96703338623047,
740
+ 21.670330047607422,
741
+ 43.46813125610358
742
+ ],
743
+ [
744
+ 16.571430206298828,
745
+ 29.010986328125,
746
+ 37.450547981262204,
747
+ 24.65054893493665,
748
+ 48.61538105010986
749
+ ],
750
+ [
751
+ 18.41758155822754,
752
+ 32.00000171661377,
753
+ 41.758243560791016,
754
+ 27.599999427795474,
755
+ 53.49011116027838
756
+ ],
757
+ [
758
+ 20.087913513183594,
759
+ 34.901100158691406,
760
+ 45.71428680419922,
761
+ 30.285715961456297,
762
+ 57.621977615356506
763
+ ],
764
+ [
765
+ 21.934066772460938,
766
+ 37.626373291015625,
767
+ 49.53406620025641,
768
+ 32.835164833068845,
769
+ 61.802198219299314
770
+ ],
771
+ [
772
+ 23.60439577102661,
773
+ 40.26373462677002,
774
+ 53.09890365600586,
775
+ 34.94505310058594,
776
+ 65.4065933227539
777
+ ],
778
+ [
779
+ 25.186813354492188,
780
+ 42.46154022216797,
781
+ 56.56703453063971,
782
+ 36.967035102844235,
783
+ 68.4835163116455
784
+ ],
785
+ [
786
+ 26.545056343078677,
787
+ 44.835164833068845,
788
+ 59.64395408630377,
789
+ 38.901100158691406,
790
+ 71.16043777465826
791
+ ],
792
+ [
793
+ 27.736264133453368,
794
+ 47.0329683303833,
795
+ 62.5054931640625,
796
+ 40.395606994628906,
797
+ 73.5824161529541
798
+ ],
799
+ [
800
+ 29.18241872787482,
801
+ 49.0549430847168,
802
+ 65.14286041259766,
803
+ 41.62637500762939,
804
+ 75.64395828247076
805
+ ],
806
+ [
807
+ 30.32527532577521,
808
+ 50.989007759094235,
809
+ 67.16483306884766,
810
+ 42.59340839385986,
811
+ 77.62637329101562
812
+ ],
813
+ [
814
+ 31.60439682006836,
815
+ 52.61098957061774,
816
+ 69.75384025573737,
817
+ 43.56044006347656,
818
+ 79.03296661376953
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:ff23bce9a8343379505324f776c39fac3ab056cc214ccb13f064235013b7429c
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 sha256:6c1fcd538566e60beb9343ec7f04592130dae8db5e2fb87e0b6d337246485290
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_insert", "run_id": "groot_tool_insert"}