viis0002 commited on
Commit
a7d726c
·
verified ·
1 Parent(s): 28e67dc

Upload folder using huggingface_hub

Browse files
checkpoint-10000/config.json ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_horizon": 50,
3
+ "add_pos_embed": true,
4
+ "apply_sincos_state_encoding": true,
5
+ "architectures": [
6
+ "Gr00tN1d6"
7
+ ],
8
+ "attn_dropout": 0.2,
9
+ "attn_implementation": null,
10
+ "backbone_embedding_dim": 2048,
11
+ "backbone_model_type": "eagle",
12
+ "backbone_trainable_params_fp32": true,
13
+ "collator_overwrite_image_inputs": false,
14
+ "color_jitter_params": {
15
+ "brightness": 0.1,
16
+ "contrast": 0.1,
17
+ "hue": 0.1,
18
+ "saturation": 0.1
19
+ },
20
+ "crop_fraction": 0.95,
21
+ "diffusion_model_cfg": {
22
+ "attention_head_dim": 48,
23
+ "dropout": 0.2,
24
+ "final_dropout": true,
25
+ "interleave_self_attention": true,
26
+ "norm_type": "ada_norm",
27
+ "num_attention_heads": 32,
28
+ "num_layers": 32,
29
+ "output_dim": 1024,
30
+ "positional_embeddings": null
31
+ },
32
+ "eagle_collator": true,
33
+ "formalize_language": true,
34
+ "gemma_collator": false,
35
+ "hidden_size": 1024,
36
+ "image_crop_size": null,
37
+ "image_target_size": null,
38
+ "input_embedding_dim": 1536,
39
+ "load_bf16": true,
40
+ "max_action_dim": 128,
41
+ "max_num_embodiments": 32,
42
+ "max_seq_len": 1024,
43
+ "max_state_dim": 128,
44
+ "model_dtype": "bfloat16",
45
+ "model_name": "nvidia/Eagle-Block2A-2B-v2",
46
+ "model_type": "Gr00tN1d6",
47
+ "noise_beta_alpha": 1.5,
48
+ "noise_beta_beta": 1.0,
49
+ "noise_s": 0.999,
50
+ "num_inference_timesteps": 4,
51
+ "num_timestep_buckets": 1000,
52
+ "random_rotation_angle": null,
53
+ "reproject_vision": false,
54
+ "select_layer": 16,
55
+ "shortest_image_edge": 256,
56
+ "state_dropout_prob": 0.0,
57
+ "torch_dtype": "bfloat16",
58
+ "transformers_version": "4.51.3",
59
+ "tune_diffusion_model": true,
60
+ "tune_llm": false,
61
+ "tune_projector": true,
62
+ "tune_top_llm_layers": 4,
63
+ "tune_visual": false,
64
+ "tune_vlln": true,
65
+ "use_albumentations_transforms": true,
66
+ "use_alternate_vl_dit": true,
67
+ "use_flash_attention": true,
68
+ "use_relative_action": true,
69
+ "use_vlln": true
70
+ }
checkpoint-10000/embodiment_id.json ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "robocasa_panda_omron": 13,
3
+ "gr1": 20,
4
+ "behavior_r1_pro": 24,
5
+ "unitree_g1": 8,
6
+ "oxe_google": 0,
7
+ "oxe_widowx": 1,
8
+ "libero_panda": 2,
9
+ "oxe_droid": 16,
10
+ "new_embodiment": 10
11
+ }
checkpoint-10000/experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,211 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ load_config_path: null
2
+ model:
3
+ model_type: Gr00tN1d6
4
+ model_dtype: bfloat16
5
+ model_name: nvidia/Eagle-Block2A-2B-v2
6
+ backbone_model_type: eagle
7
+ model_revision: null
8
+ tune_top_llm_layers: 4
9
+ backbone_embedding_dim: 2048
10
+ tune_llm: false
11
+ tune_visual: false
12
+ select_layer: 16
13
+ reproject_vision: false
14
+ use_flash_attention: true
15
+ load_bf16: false
16
+ collator_overwrite_image_inputs: false
17
+ eagle_collator: true
18
+ backbone_trainable_params_fp32: true
19
+ image_crop_size: null
20
+ image_target_size: null
21
+ shortest_image_edge: 256
22
+ crop_fraction: 0.95
23
+ random_rotation_angle: null
24
+ color_jitter_params:
25
+ brightness: 0.3
26
+ contrast: 0.4
27
+ saturation: 0.5
28
+ hue: 0.08
29
+ use_albumentations_transforms: true
30
+ extra_augmentation_config: null
31
+ formalize_language: true
32
+ apply_sincos_state_encoding: false
33
+ use_relative_action: true
34
+ max_state_dim: 29
35
+ max_action_dim: 29
36
+ action_horizon: 16
37
+ hidden_size: 1024
38
+ input_embedding_dim: 1536
39
+ add_pos_embed: true
40
+ attn_dropout: 0.2
41
+ use_vlln: true
42
+ max_seq_len: 1024
43
+ use_alternate_vl_dit: true
44
+ attend_text_every_n_blocks: 2
45
+ diffusion_model_cfg:
46
+ positional_embeddings: null
47
+ num_layers: 32
48
+ num_attention_heads: 32
49
+ attention_head_dim: 48
50
+ norm_type: ada_norm
51
+ dropout: 0.2
52
+ final_dropout: true
53
+ output_dim: 1024
54
+ interleave_self_attention: true
55
+ num_inference_timesteps: 4
56
+ noise_beta_alpha: 1.5
57
+ noise_beta_beta: 1.0
58
+ noise_s: 0.999
59
+ num_timestep_buckets: 1000
60
+ tune_projector: true
61
+ tune_diffusion_model: true
62
+ tune_vlln: true
63
+ state_dropout_prob: 0.0
64
+ state_additive_noise_scale: 0.0
65
+ max_num_embodiments: 32
66
+ data:
67
+ datasets:
68
+ - dataset_paths:
69
+ - ./demo_data/ps2/viis0002/cluttered_picking_v2
70
+ embodiment_tag: new_embodiment
71
+ mix_ratio: 1.0
72
+ dataset_type: physical_embodiment
73
+ val_dataset_path: null
74
+ modality_configs:
75
+ new_embodiment:
76
+ video:
77
+ delta_indices:
78
+ - 0
79
+ modality_keys:
80
+ - front
81
+ - wrist_color
82
+ - wrist_depth
83
+ sin_cos_embedding_keys: null
84
+ mean_std_embedding_keys: null
85
+ action_configs: null
86
+ state:
87
+ delta_indices:
88
+ - 0
89
+ modality_keys:
90
+ - single_arm
91
+ - gripper
92
+ sin_cos_embedding_keys: null
93
+ mean_std_embedding_keys: null
94
+ action_configs: null
95
+ action:
96
+ delta_indices:
97
+ - 0
98
+ - 1
99
+ - 2
100
+ - 3
101
+ - 4
102
+ - 5
103
+ - 6
104
+ - 7
105
+ - 8
106
+ - 9
107
+ - 10
108
+ - 11
109
+ - 12
110
+ - 13
111
+ - 14
112
+ - 15
113
+ modality_keys:
114
+ - single_arm
115
+ - gripper
116
+ sin_cos_embedding_keys: null
117
+ mean_std_embedding_keys: null
118
+ action_configs:
119
+ - rep: RELATIVE
120
+ type: NON_EEF
121
+ format: DEFAULT
122
+ state_key: null
123
+ - rep: ABSOLUTE
124
+ type: NON_EEF
125
+ format: DEFAULT
126
+ state_key: null
127
+ language:
128
+ delta_indices:
129
+ - 0
130
+ modality_keys:
131
+ - annotation.human.task_description
132
+ sin_cos_embedding_keys: null
133
+ mean_std_embedding_keys: null
134
+ action_configs: null
135
+ download_cache: false
136
+ shard_size: 1024
137
+ episode_sampling_rate: 0.1
138
+ num_shards_per_epoch: 100000
139
+ override_pretraining_statistics: false
140
+ mode: single_turn
141
+ random_chop: 0.0
142
+ mock_dataset_mode: false
143
+ shuffle: true
144
+ seed: 42
145
+ multiprocessing_context: fork
146
+ allow_padding: false
147
+ subsample_ratio: 1.0
148
+ image_crop_size:
149
+ - 244
150
+ - 244
151
+ image_target_size:
152
+ - 224
153
+ - 224
154
+ video_backend: torchcodec
155
+ training:
156
+ output_dir: ./output
157
+ experiment_name: null
158
+ max_steps: 10000
159
+ global_batch_size: 8
160
+ batch_size: null
161
+ gradient_accumulation_steps: 1
162
+ learning_rate: 0.0001
163
+ lr_scheduler_type: cosine
164
+ weight_decay: 1.0e-05
165
+ warmup_ratio: 0.05
166
+ warmup_steps: 0
167
+ max_grad_norm: 1.0
168
+ optim: adamw_torch
169
+ start_from_checkpoint: nvidia/GR00T-N1.6-3B
170
+ tf32: true
171
+ fp16: false
172
+ bf16: true
173
+ eval_bf16: true
174
+ logging_steps: 10
175
+ save_steps: 3000
176
+ save_total_limit: 5
177
+ save_vl_model: false
178
+ upload_checkpoints: false
179
+ upload_every: 1000
180
+ upload_last_n_checkpoints: 5
181
+ max_concurrent_uploads: 2
182
+ eval_strategy: 'no'
183
+ eval_steps: 500
184
+ eval_set_split_ratio: 0.1
185
+ eval_batch_size: 2
186
+ save_best_eval_metric_name: ''
187
+ save_best_eval_metric_greater_is_better: true
188
+ deepspeed_stage: 2
189
+ gradient_checkpointing: false
190
+ transformers_trust_remote_code: true
191
+ transformers_local_files_only: false
192
+ transformers_cache_dir: null
193
+ transformers_access_token: null
194
+ use_ddp: false
195
+ ddp_bucket_cap_mb: 100
196
+ num_gpus: 1
197
+ dataloader_num_workers: 2
198
+ remove_unused_columns: false
199
+ use_wandb: false
200
+ wandb_project: finetune-gr00t-n1d6
201
+ enable_profiling: false
202
+ max_retries: 3
203
+ assert_loss_less_than: null
204
+ add_rl_callback: false
205
+ enable_open_loop_eval: false
206
+ open_loop_eval_traj_ids:
207
+ - 0
208
+ open_loop_eval_steps_per_traj: 100
209
+ open_loop_eval_plot_indices: null
210
+ max_steps: 10000
211
+ save_steps: 3000
checkpoint-10000/experiment_cfg/config.yaml ADDED
@@ -0,0 +1,245 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ !!python/object:gr00t.configs.base_config.Config
2
+ data: !!python/object:gr00t.configs.data.data_config.DataConfig
3
+ allow_padding: false
4
+ datasets:
5
+ - !!python/object:gr00t.configs.data.data_config.SingleDatasetConfig
6
+ dataset_paths:
7
+ - ./demo_data/ps2/viis0002/cluttered_picking_v2
8
+ dataset_type: physical_embodiment
9
+ embodiment_tag: new_embodiment
10
+ mix_ratio: 1.0
11
+ val_dataset_path: null
12
+ download_cache: false
13
+ episode_sampling_rate: 0.1
14
+ image_crop_size:
15
+ - 244
16
+ - 244
17
+ image_target_size:
18
+ - 224
19
+ - 224
20
+ mock_dataset_mode: false
21
+ modality_configs:
22
+ new_embodiment:
23
+ action: !!python/object:gr00t.data.types.ModalityConfig
24
+ action_configs:
25
+ - !!python/object:gr00t.data.types.ActionConfig
26
+ format: &id001 !!python/object/apply:gr00t.data.types.ActionFormat
27
+ - default
28
+ rep: !!python/object/apply:gr00t.data.types.ActionRepresentation
29
+ - relative
30
+ state_key: null
31
+ type: &id002 !!python/object/apply:gr00t.data.types.ActionType
32
+ - non_eef
33
+ - !!python/object:gr00t.data.types.ActionConfig
34
+ format: *id001
35
+ rep: !!python/object/apply:gr00t.data.types.ActionRepresentation
36
+ - absolute
37
+ state_key: null
38
+ type: *id002
39
+ delta_indices:
40
+ - 0
41
+ - 1
42
+ - 2
43
+ - 3
44
+ - 4
45
+ - 5
46
+ - 6
47
+ - 7
48
+ - 8
49
+ - 9
50
+ - 10
51
+ - 11
52
+ - 12
53
+ - 13
54
+ - 14
55
+ - 15
56
+ mean_std_embedding_keys: null
57
+ modality_keys:
58
+ - single_arm
59
+ - gripper
60
+ sin_cos_embedding_keys: null
61
+ language: !!python/object:gr00t.data.types.ModalityConfig
62
+ action_configs: null
63
+ delta_indices:
64
+ - 0
65
+ mean_std_embedding_keys: null
66
+ modality_keys:
67
+ - annotation.human.task_description
68
+ sin_cos_embedding_keys: null
69
+ state: !!python/object:gr00t.data.types.ModalityConfig
70
+ action_configs: null
71
+ delta_indices:
72
+ - 0
73
+ mean_std_embedding_keys: null
74
+ modality_keys:
75
+ - single_arm
76
+ - gripper
77
+ sin_cos_embedding_keys: null
78
+ video: !!python/object:gr00t.data.types.ModalityConfig
79
+ action_configs: null
80
+ delta_indices:
81
+ - 0
82
+ mean_std_embedding_keys: null
83
+ modality_keys:
84
+ - front
85
+ - wrist_color
86
+ - wrist_depth
87
+ sin_cos_embedding_keys: null
88
+ mode: single_turn
89
+ multiprocessing_context: fork
90
+ num_shards_per_epoch: 100000
91
+ override_pretraining_statistics: false
92
+ random_chop: 0.0
93
+ seed: 42
94
+ shard_size: 1024
95
+ shuffle: true
96
+ subsample_ratio: 1.0
97
+ video_backend: torchcodec
98
+ load_config_path: null
99
+ model: !!python/object:gr00t.configs.model.gr00t_n1d6.Gr00tN1d6Config
100
+ _attn_implementation_autoset: false
101
+ _attn_implementation_internal: null
102
+ _commit_hash: null
103
+ _name_or_path: ''
104
+ add_cross_attention: false
105
+ architectures: null
106
+ backbone_model_type: eagle
107
+ backbone_trainable_params_fp32: true
108
+ bad_words_ids: null
109
+ begin_suppress_tokens: null
110
+ bos_token_id: null
111
+ chunk_size_feed_forward: 0
112
+ color_jitter_params:
113
+ brightness: 0.3
114
+ contrast: 0.4
115
+ hue: 0.08
116
+ saturation: 0.5
117
+ cross_attention_hidden_size: null
118
+ decoder_start_token_id: null
119
+ diffusion_model_cfg:
120
+ attention_head_dim: 48
121
+ dropout: 0.2
122
+ final_dropout: true
123
+ interleave_self_attention: true
124
+ norm_type: ada_norm
125
+ num_attention_heads: 32
126
+ num_layers: 32
127
+ output_dim: 1024
128
+ positional_embeddings: null
129
+ diversity_penalty: 0.0
130
+ do_sample: false
131
+ eagle_collator: true
132
+ early_stopping: false
133
+ encoder_no_repeat_ngram_size: 0
134
+ eos_token_id: null
135
+ exponential_decay_length_penalty: null
136
+ extra_augmentation_config: null
137
+ finetuning_task: null
138
+ forced_bos_token_id: null
139
+ forced_eos_token_id: null
140
+ id2label:
141
+ 0: LABEL_0
142
+ 1: LABEL_1
143
+ is_decoder: false
144
+ is_encoder_decoder: false
145
+ label2id:
146
+ LABEL_0: 0
147
+ LABEL_1: 1
148
+ length_penalty: 1.0
149
+ load_bf16: false
150
+ max_length: 20
151
+ min_length: 0
152
+ model_name: nvidia/Eagle-Block2A-2B-v2
153
+ no_repeat_ngram_size: 0
154
+ num_beam_groups: 1
155
+ num_beams: 1
156
+ num_return_sequences: 1
157
+ output_attentions: false
158
+ output_hidden_states: false
159
+ output_scores: false
160
+ pad_token_id: null
161
+ prefix: null
162
+ problem_type: null
163
+ pruned_heads: {}
164
+ random_rotation_angle: null
165
+ remove_invalid_values: false
166
+ repetition_penalty: 1.0
167
+ reproject_vision: false
168
+ return_dict: true
169
+ return_dict_in_generate: false
170
+ sep_token_id: null
171
+ state_dropout_prob: 0.0
172
+ suppress_tokens: null
173
+ task_specific_params: null
174
+ temperature: 1.0
175
+ tf_legacy_loss: false
176
+ tie_encoder_decoder: false
177
+ tie_word_embeddings: true
178
+ tokenizer_class: null
179
+ top_k: 50
180
+ top_p: 1.0
181
+ torch_dtype: null
182
+ torchscript: false
183
+ transformers_version: null
184
+ tune_diffusion_model: true
185
+ tune_llm: false
186
+ tune_projector: true
187
+ tune_visual: false
188
+ typical_p: 1.0
189
+ use_bfloat16: false
190
+ use_relative_action: true
191
+ training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
192
+ add_rl_callback: false
193
+ assert_loss_less_than: null
194
+ batch_size: null
195
+ bf16: true
196
+ dataloader_num_workers: 2
197
+ ddp_bucket_cap_mb: 100
198
+ deepspeed_stage: 2
199
+ enable_open_loop_eval: false
200
+ enable_profiling: false
201
+ eval_batch_size: 2
202
+ eval_bf16: true
203
+ eval_set_split_ratio: 0.1
204
+ eval_steps: 500
205
+ eval_strategy: 'no'
206
+ experiment_name: null
207
+ fp16: false
208
+ global_batch_size: 8
209
+ gradient_accumulation_steps: 1
210
+ gradient_checkpointing: false
211
+ learning_rate: 0.0001
212
+ logging_steps: 10
213
+ lr_scheduler_type: cosine
214
+ max_concurrent_uploads: 2
215
+ max_grad_norm: 1.0
216
+ max_retries: 3
217
+ max_steps: 10000
218
+ num_gpus: 1
219
+ open_loop_eval_plot_indices: null
220
+ open_loop_eval_steps_per_traj: 100
221
+ open_loop_eval_traj_ids:
222
+ - 0
223
+ optim: adamw_torch
224
+ output_dir: ./output
225
+ remove_unused_columns: false
226
+ save_best_eval_metric_greater_is_better: true
227
+ save_best_eval_metric_name: ''
228
+ save_steps: 3000
229
+ save_total_limit: 5
230
+ save_vl_model: false
231
+ start_from_checkpoint: nvidia/GR00T-N1.6-3B
232
+ tf32: true
233
+ transformers_access_token: null
234
+ transformers_cache_dir: null
235
+ transformers_local_files_only: false
236
+ transformers_trust_remote_code: true
237
+ upload_checkpoints: false
238
+ upload_every: 1000
239
+ upload_last_n_checkpoints: 5
240
+ use_ddp: false
241
+ use_wandb: false
242
+ wandb_project: finetune-gr00t-n1d6
243
+ warmup_ratio: 0.05
244
+ warmup_steps: 0
245
+ weight_decay: 1.0e-05
checkpoint-10000/experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,824 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "single_arm": {
5
+ "min": [
6
+ -43.319149017333984,
7
+ -99.40652465820312,
8
+ -39.10285568237305,
9
+ 48.36065673828125,
10
+ -34.70085525512695
11
+ ],
12
+ "max": [
13
+ 11.829787254333496,
14
+ 40.398475646972656,
15
+ 100.0,
16
+ 100.0,
17
+ 56.43467712402344
18
+ ],
19
+ "mean": [
20
+ -4.440749271172309,
21
+ -13.308821339499419,
22
+ 15.1325215701596,
23
+ 94.23170722591506,
24
+ 9.425153629722358
25
+ ],
26
+ "std": [
27
+ 12.14406229008906,
28
+ 36.99829120476212,
29
+ 34.39080809147234,
30
+ 9.257896912650017,
31
+ 16.60771267620677
32
+ ],
33
+ "q01": [
34
+ -32.765687240006784,
35
+ -98.18120676093605,
36
+ -20.29557584031693,
37
+ 67.62860396622403,
38
+ -5.580963820366954
39
+ ],
40
+ "q99": [
41
+ 8.45665083715747,
42
+ 31.670184474024747,
43
+ 95.99070525799908,
44
+ 99.69872470853615,
45
+ 40.65240490687968
46
+ ]
47
+ },
48
+ "gripper": {
49
+ "min": [
50
+ 0.5750798583030701
51
+ ],
52
+ "max": [
53
+ 73.9297103881836
54
+ ],
55
+ "mean": [
56
+ 9.998555719102907
57
+ ],
58
+ "std": [
59
+ 12.050090684502234
60
+ ],
61
+ "q01": [
62
+ 1.197928574688387
63
+ ],
64
+ "q99": [
65
+ 47.81798753608932
66
+ ]
67
+ }
68
+ },
69
+ "action": {
70
+ "single_arm": {
71
+ "min": [
72
+ -43.884891510009766,
73
+ -99.5762710571289,
74
+ -43.259193420410156,
75
+ 46.74245834350586,
76
+ -34.94505310058594
77
+ ],
78
+ "max": [
79
+ 12.53313159942627,
80
+ 41.949153900146484,
81
+ 99.63685607910156,
82
+ 100.0,
83
+ 56.82539749145508
84
+ ],
85
+ "mean": [
86
+ -4.336263196603688,
87
+ -15.522390066565956,
88
+ 13.258741117242543,
89
+ 94.06182767417421,
90
+ 9.397664254783678
91
+ ],
92
+ "std": [
93
+ 12.372809314706336,
94
+ 36.56167326493358,
95
+ 34.92591755064374,
96
+ 9.370667724934858,
97
+ 16.699527746233734
98
+ ],
99
+ "q01": [
100
+ -33.256299413939324,
101
+ -98.9770363306071,
102
+ -23.248310714347284,
103
+ 66.50771469329445,
104
+ -5.869576220557871
105
+ ],
106
+ "q99": [
107
+ 8.985025713499576,
108
+ 31.601930304179678,
109
+ 93.41583007241056,
110
+ 99.88448441811833,
111
+ 40.89866939535895
112
+ ]
113
+ },
114
+ "gripper": {
115
+ "min": [
116
+ 0.0
117
+ ],
118
+ "max": [
119
+ 74.22003173828125
120
+ ],
121
+ "mean": [
122
+ 7.439182423450768
123
+ ],
124
+ "std": [
125
+ 12.415960562905878
126
+ ],
127
+ "q01": [
128
+ 0.7503314920264533
129
+ ],
130
+ "q99": [
131
+ 48.41172194516817
132
+ ]
133
+ }
134
+ },
135
+ "relative_action": {
136
+ "single_arm": {
137
+ "min": [
138
+ [
139
+ -14.407760620117188,
140
+ -27.423320770263672,
141
+ -21.879257202148438,
142
+ -48.30026626586914,
143
+ -45.86080551147461
144
+ ],
145
+ [
146
+ -14.407760620117188,
147
+ -27.423320770263672,
148
+ -22.60553741455078,
149
+ -48.30026626586914,
150
+ -45.86080551147461
151
+ ],
152
+ [
153
+ -14.407760620117188,
154
+ -27.423320770263672,
155
+ -25.817413330078125,
156
+ -48.30026626586914,
157
+ -45.86080551147461
158
+ ],
159
+ [
160
+ -15.768430709838867,
161
+ -27.423320770263672,
162
+ -29.261001586914062,
163
+ -48.30026626586914,
164
+ -45.86080551147461
165
+ ],
166
+ [
167
+ -17.888839721679688,
168
+ -27.423320770263672,
169
+ -32.529273986816406,
170
+ -48.30026626586914,
171
+ -45.86080551147461
172
+ ],
173
+ [
174
+ -19.93351936340332,
175
+ -27.423320770263672,
176
+ -35.610374450683594,
177
+ -48.30026626586914,
178
+ -45.86080551147461
179
+ ],
180
+ [
181
+ -21.82674217224121,
182
+ -27.68240737915039,
183
+ -38.79345703125,
184
+ -48.30026626586914,
185
+ -45.86080551147461
186
+ ],
187
+ [
188
+ -23.699081420898438,
189
+ -29.25457000732422,
190
+ -41.87455749511719,
191
+ -48.30026626586914,
192
+ -45.86080551147461
193
+ ],
194
+ [
195
+ -25.51657485961914,
196
+ -31.56486701965332,
197
+ -44.69236373901367,
198
+ -48.30026626586914,
199
+ -45.86080551147461
200
+ ],
201
+ [
202
+ -27.258338928222656,
203
+ -34.022499084472656,
204
+ -47.77346420288086,
205
+ -48.30026626586914,
206
+ -45.86080551147461
207
+ ],
208
+ [
209
+ -28.772918701171875,
210
+ -36.31063461303711,
211
+ -50.587806701660156,
212
+ -48.30026626586914,
213
+ -45.86080551147461
214
+ ],
215
+ [
216
+ -30.060306549072266,
217
+ -38.68351364135742,
218
+ -53.2205810546875,
219
+ -48.30026626586914,
220
+ -45.86080551147461
221
+ ],
222
+ [
223
+ -31.120513916015625,
224
+ -40.971649169921875,
225
+ -55.757965087890625,
226
+ -48.30026626586914,
227
+ -45.86080551147461
228
+ ],
229
+ [
230
+ -32.14179229736328,
231
+ -43.17504119873047,
232
+ -58.114105224609375,
233
+ -48.30026626586914,
234
+ -45.86080551147461
235
+ ],
236
+ [
237
+ -33.37945556640625,
238
+ -45.708072662353516,
239
+ -60.20216751098633,
240
+ -48.30026626586914,
241
+ -45.86080551147461
242
+ ],
243
+ [
244
+ -34.48583984375,
245
+ -48.25044631958008,
246
+ -62.19582748413086,
247
+ -48.30026626586914,
248
+ -45.86080551147461
249
+ ]
250
+ ],
251
+ "max": [
252
+ [
253
+ 10.07719612121582,
254
+ 18.55738067626953,
255
+ 13.711788177490234,
256
+ 11.358840942382812,
257
+ 11.868131637573242
258
+ ],
259
+ [
260
+ 12.046146392822266,
261
+ 20.422576904296875,
262
+ 13.711788177490234,
263
+ 14.330146789550781,
264
+ 14.163614273071289
265
+ ],
266
+ [
267
+ 13.863639831542969,
268
+ 22.11749267578125,
269
+ 16.409061431884766,
270
+ 16.994083404541016,
271
+ 16.361417770385742
272
+ ],
273
+ [
274
+ 15.792083740234375,
275
+ 24.909435272216797,
276
+ 19.24298095703125,
277
+ 19.530155181884766,
278
+ 18.168498992919922
279
+ ],
280
+ [
281
+ 17.912492752075195,
282
+ 28.13056182861328,
283
+ 22.329681396484375,
284
+ 22.153682708740234,
285
+ 19.78022003173828
286
+ ],
287
+ [
288
+ 19.7299861907959,
289
+ 31.18191146850586,
290
+ 25.320167541503906,
291
+ 24.689754486083984,
292
+ 21.098901748657227
293
+ ],
294
+ [
295
+ 21.517221450805664,
296
+ 34.232757568359375,
297
+ 28.40127182006836,
298
+ 27.05988311767578,
299
+ 23.10134506225586
300
+ ],
301
+ [
302
+ 23.25898551940918,
303
+ 36.85987854003906,
304
+ 31.215618133544922,
305
+ 29.407485961914062,
306
+ 25.10378646850586
307
+ ],
308
+ [
309
+ 24.96111297607422,
310
+ 39.402252197265625,
311
+ 33.934242248535156,
312
+ 31.68120574951172,
313
+ 27.008548736572266
314
+ ],
315
+ [
316
+ 26.32281494140625,
317
+ 41.860923767089844,
318
+ 36.5670166015625,
319
+ 33.69257354736328,
320
+ 28.913311004638672
321
+ ],
322
+ [
323
+ 27.610206604003906,
324
+ 44.233802795410156,
325
+ 39.19501876831055,
326
+ 35.63929748535156,
327
+ 30.818073272705078
328
+ ],
329
+ [
330
+ 28.746139526367188,
331
+ 46.68543243408203,
332
+ 41.737003326416016,
333
+ 37.38831329345703,
334
+ 32.625152587890625
335
+ ],
336
+ [
337
+ 29.80634307861328,
338
+ 49.737571716308594,
339
+ 44.274383544921875,
340
+ 38.92519760131836,
341
+ 34.334556579589844
342
+ ],
343
+ [
344
+ 30.57229995727539,
345
+ 52.535362243652344,
346
+ 46.634803771972656,
347
+ 40.14950942993164,
348
+ 36.141639709472656
349
+ ],
350
+ [
351
+ 31.500513076782227,
352
+ 55.078819274902344,
353
+ 48.98831558227539,
354
+ 40.76426315307617,
355
+ 37.80220031738281
356
+ ],
357
+ [
358
+ 33.01509094238281,
359
+ 57.621192932128906,
360
+ 51.16715621948242,
361
+ 41.1302375793457,
362
+ 39.120880126953125
363
+ ]
364
+ ],
365
+ "mean": [
366
+ [
367
+ 0.10973429679870605,
368
+ -2.209338903427124,
369
+ -1.8722033500671387,
370
+ -0.15508916974067688,
371
+ -0.028145914897322655
372
+ ],
373
+ [
374
+ 0.10968202352523804,
375
+ -2.2086832523345947,
376
+ -1.8730086088180542,
377
+ -0.154344841837883,
378
+ -0.028264271095395088
379
+ ],
380
+ [
381
+ 0.10962218046188354,
382
+ -2.20805287361145,
383
+ -1.873792290687561,
384
+ -0.15358231961727142,
385
+ -0.028378551825881004
386
+ ],
387
+ [
388
+ 0.10955725610256195,
389
+ -2.207461357116699,
390
+ -1.874544382095337,
391
+ -0.15280047059059143,
392
+ -0.02848830074071884
393
+ ],
394
+ [
395
+ 0.10948893427848816,
396
+ -2.2069222927093506,
397
+ -1.8752955198287964,
398
+ -0.15200501680374146,
399
+ -0.02859322726726532
400
+ ],
401
+ [
402
+ 0.10941897332668304,
403
+ -2.20639705657959,
404
+ -1.8760193586349487,
405
+ -0.15119512379169464,
406
+ -0.028693431988358498
407
+ ],
408
+ [
409
+ 0.10934928804636002,
410
+ -2.2059144973754883,
411
+ -1.8767318725585938,
412
+ -0.1503802090883255,
413
+ -0.028788452968001366
414
+ ],
415
+ [
416
+ 0.1092788428068161,
417
+ -2.205451726913452,
418
+ -1.8774136304855347,
419
+ -0.1495615541934967,
420
+ -0.02887848950922489
421
+ ],
422
+ [
423
+ 0.109209343791008,
424
+ -2.2050089836120605,
425
+ -1.8780895471572876,
426
+ -0.1487341672182083,
427
+ -0.028965836390852928
428
+ ],
429
+ [
430
+ 0.10913944989442825,
431
+ -2.2045910358428955,
432
+ -1.8787379264831543,
433
+ -0.14790278673171997,
434
+ -0.02905186079442501
435
+ ],
436
+ [
437
+ 0.10906971991062164,
438
+ -2.2042009830474854,
439
+ -1.8793729543685913,
440
+ -0.14706464111804962,
441
+ -0.02913922630250454
442
+ ],
443
+ [
444
+ 0.10899965465068817,
445
+ -2.2038326263427734,
446
+ -1.879947304725647,
447
+ -0.14621910452842712,
448
+ -0.029229290783405304
449
+ ],
450
+ [
451
+ 0.10893117636442184,
452
+ -2.203528881072998,
453
+ -1.8804959058761597,
454
+ -0.14536546170711517,
455
+ -0.029323536902666092
456
+ ],
457
+ [
458
+ 0.10886196792125702,
459
+ -2.2032315731048584,
460
+ -1.8810045719146729,
461
+ -0.14450465142726898,
462
+ -0.029420997947454453
463
+ ],
464
+ [
465
+ 0.1087912917137146,
466
+ -2.2029733657836914,
467
+ -1.8814704418182373,
468
+ -0.1436382681131363,
469
+ -0.029520653188228607
470
+ ],
471
+ [
472
+ 0.10871820151805878,
473
+ -2.2027506828308105,
474
+ -1.8818960189819336,
475
+ -0.14276832342147827,
476
+ -0.029622290283441544
477
+ ]
478
+ ],
479
+ "std": [
480
+ [
481
+ 1.291955590248108,
482
+ 2.895820140838623,
483
+ 2.4284205436706543,
484
+ 1.0855623483657837,
485
+ 0.9884634017944336
486
+ ],
487
+ [
488
+ 1.5667155981063843,
489
+ 3.248924732208252,
490
+ 2.7988362312316895,
491
+ 1.2348947525024414,
492
+ 1.1902235746383667
493
+ ],
494
+ [
495
+ 1.8442904949188232,
496
+ 3.62117600440979,
497
+ 3.1847639083862305,
498
+ 1.3908764123916626,
499
+ 1.3953996896743774
500
+ ],
501
+ [
502
+ 2.121694564819336,
503
+ 4.004567623138428,
504
+ 3.5781822204589844,
505
+ 1.5495537519454956,
506
+ 1.6011954545974731
507
+ ],
508
+ [
509
+ 2.397252321243286,
510
+ 4.3943281173706055,
511
+ 3.9742770195007324,
512
+ 1.7082926034927368,
513
+ 1.8060038089752197
514
+ ],
515
+ [
516
+ 2.6699206829071045,
517
+ 4.787065505981445,
518
+ 4.36962890625,
519
+ 1.8654924631118774,
520
+ 2.009084939956665
521
+ ],
522
+ [
523
+ 2.9392967224121094,
524
+ 5.180052757263184,
525
+ 4.762381553649902,
526
+ 2.0205588340759277,
527
+ 2.209796190261841
528
+ ],
529
+ [
530
+ 3.2050225734710693,
531
+ 5.571920871734619,
532
+ 5.15108060836792,
533
+ 2.1726198196411133,
534
+ 2.4078612327575684
535
+ ],
536
+ [
537
+ 3.466845750808716,
538
+ 5.961581707000732,
539
+ 5.5352396965026855,
540
+ 2.3211541175842285,
541
+ 2.602936267852783
542
+ ],
543
+ [
544
+ 3.724405527114868,
545
+ 6.348147869110107,
546
+ 5.91373348236084,
547
+ 2.4658448696136475,
548
+ 2.7950327396392822
549
+ ],
550
+ [
551
+ 3.977760076522827,
552
+ 6.730666160583496,
553
+ 6.286655902862549,
554
+ 2.606712579727173,
555
+ 2.983889579772949
556
+ ],
557
+ [
558
+ 4.226820468902588,
559
+ 7.108645915985107,
560
+ 6.653498649597168,
561
+ 2.7434158325195312,
562
+ 3.169560194015503
563
+ ],
564
+ [
565
+ 4.471673965454102,
566
+ 7.482038974761963,
567
+ 7.014378070831299,
568
+ 2.8758957386016846,
569
+ 3.3519320487976074
570
+ ],
571
+ [
572
+ 4.7121500968933105,
573
+ 7.85037899017334,
574
+ 7.368745803833008,
575
+ 3.004105567932129,
576
+ 3.5309762954711914
577
+ ],
578
+ [
579
+ 4.9483323097229,
580
+ 8.213367462158203,
581
+ 7.716771602630615,
582
+ 3.1281301975250244,
583
+ 3.7066855430603027
584
+ ],
585
+ [
586
+ 5.179955005645752,
587
+ 8.571035385131836,
588
+ 8.058523178100586,
589
+ 3.248170852661133,
590
+ 3.879103422164917
591
+ ]
592
+ ],
593
+ "q01": [
594
+ [
595
+ -3.880250930786133,
596
+ -8.599881172180176,
597
+ -9.154699325561523,
598
+ -3.436232089996338,
599
+ -3.425642251968384
600
+ ],
601
+ [
602
+ -4.857089996337891,
603
+ -9.876320838928223,
604
+ -10.865798950195312,
605
+ -3.8265511989593506,
606
+ -4.346764087677002
607
+ ],
608
+ [
609
+ -5.876626968383789,
610
+ -11.200237274169922,
611
+ -12.55186939239502,
612
+ -4.3084259033203125,
613
+ -5.274723052978516
614
+ ],
615
+ [
616
+ -6.864691734313965,
617
+ -12.497199058532715,
618
+ -14.240229606628418,
619
+ -4.79156494140625,
620
+ -6.153846263885498
621
+ ],
622
+ [
623
+ -7.865577220916748,
624
+ -13.833037376403809,
625
+ -15.899080276489258,
626
+ -5.358405590057373,
627
+ -7.0329670906066895
628
+ ],
629
+ [
630
+ -8.856843948364258,
631
+ -15.103858947753906,
632
+ -17.418010711669922,
633
+ -5.889519214630127,
634
+ -7.912088394165039
635
+ ],
636
+ [
637
+ -9.819734573364258,
638
+ -16.37712860107422,
639
+ -18.992219924926758,
640
+ -6.4837646484375,
641
+ -8.791208267211914
642
+ ],
643
+ [
644
+ -10.775996208190918,
645
+ -17.66802978515625,
646
+ -20.44565200805664,
647
+ -6.9591569900512695,
648
+ -9.670330047607422
649
+ ],
650
+ [
651
+ -11.732026100158691,
652
+ -18.8674259185791,
653
+ -21.919971466064453,
654
+ -7.49139928817749,
655
+ -10.556291580200195
656
+ ],
657
+ [
658
+ -12.65311050415039,
659
+ -20.0970516204834,
660
+ -23.280298233032227,
661
+ -7.985642910003662,
662
+ -11.428571701049805
663
+ ],
664
+ [
665
+ -13.55053997039795,
666
+ -21.310543060302734,
667
+ -24.66961669921875,
668
+ -8.45219612121582,
669
+ -12.258852005004883
670
+ ],
671
+ [
672
+ -14.445900917053223,
673
+ -22.47014617919922,
674
+ -25.971336364746094,
675
+ -8.819541931152344,
676
+ -13.089134216308594
677
+ ],
678
+ [
679
+ -15.32150650024414,
680
+ -23.607933044433594,
681
+ -27.2856502532959,
682
+ -9.188517570495605,
683
+ -13.919414520263672
684
+ ],
685
+ [
686
+ -16.171899795532227,
687
+ -24.695764541625977,
688
+ -28.55253028869629,
689
+ -9.590324401855469,
690
+ -14.700855255126953
691
+ ],
692
+ [
693
+ -17.013334274291992,
694
+ -25.719482421875,
695
+ -29.74740219116211,
696
+ -9.954161643981934,
697
+ -15.531135559082031
698
+ ],
699
+ [
700
+ -17.83913803100586,
701
+ -26.79294204711914,
702
+ -30.886316299438477,
703
+ -10.289176940917969,
704
+ -16.263734817504883
705
+ ]
706
+ ],
707
+ "q99": [
708
+ [
709
+ 4.964918613433838,
710
+ 5.23189640045166,
711
+ 5.2053070068359375,
712
+ 2.6407511234283447,
713
+ 3.9072036743164062
714
+ ],
715
+ [
716
+ 6.182013511657715,
717
+ 6.905364513397217,
718
+ 6.732197284698486,
719
+ 3.2062454223632812,
720
+ 4.9328460693359375
721
+ ],
722
+ [
723
+ 7.405695915222168,
724
+ 8.520384788513184,
725
+ 8.360859870910645,
726
+ 3.5569162368774414,
727
+ 5.958484649658203
728
+ ],
729
+ [
730
+ 8.610760688781738,
731
+ 10.048929214477539,
732
+ 9.90613842010498,
733
+ 4.091148376464844,
734
+ 6.89331579208374
735
+ ],
736
+ [
737
+ 9.801132202148438,
738
+ 11.651291847229004,
739
+ 11.436395645141602,
740
+ 4.566997051239014,
741
+ 7.863248348236084
742
+ ],
743
+ [
744
+ 10.974348068237305,
745
+ 13.181883811950684,
746
+ 12.88714599609375,
747
+ 5.0329437255859375,
748
+ 8.79120922088623
749
+ ],
750
+ [
751
+ 12.129141807556152,
752
+ 14.657190322875977,
753
+ 14.390020370483398,
754
+ 5.57182502746582,
755
+ 9.768009185791016
756
+ ],
757
+ [
758
+ 13.220808029174805,
759
+ 16.158241271972656,
760
+ 15.842169761657715,
761
+ 6.1076812744140625,
762
+ 10.647130966186523
763
+ ],
764
+ [
765
+ 14.329455375671387,
766
+ 17.674537658691406,
767
+ 17.28555679321289,
768
+ 6.6834516525268555,
769
+ 11.526250839233398
770
+ ],
771
+ [
772
+ 15.360580444335938,
773
+ 19.114492416381836,
774
+ 18.61423110961914,
775
+ 7.178110599517822,
776
+ 12.45421314239502
777
+ ],
778
+ [
779
+ 16.411075592041016,
780
+ 20.48096466064453,
781
+ 19.966537475585938,
782
+ 7.684507846832275,
783
+ 13.333332061767578
784
+ ],
785
+ [
786
+ 17.393070220947266,
787
+ 21.859176635742188,
788
+ 21.3115291595459,
789
+ 8.17349910736084,
790
+ 14.261293411254883
791
+ ],
792
+ [
793
+ 18.34865951538086,
794
+ 23.18575096130371,
795
+ 22.599977493286133,
796
+ 8.620201110839844,
797
+ 15.042734146118164
798
+ ],
799
+ [
800
+ 19.292339324951172,
801
+ 24.476245880126953,
802
+ 23.886260986328125,
803
+ 8.993049621582031,
804
+ 15.921854972839355
805
+ ],
806
+ [
807
+ 20.147075653076172,
808
+ 25.74358367919922,
809
+ 25.125085830688477,
810
+ 9.504680633544922,
811
+ 16.703296661376953
812
+ ],
813
+ [
814
+ 21.00503158569336,
815
+ 26.99942398071289,
816
+ 26.240026473999023,
817
+ 9.915670394897461,
818
+ 17.53357696533203
819
+ ]
820
+ ]
821
+ }
822
+ }
823
+ }
824
+ }
checkpoint-10000/experiment_cfg/final_model_config.json ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_type": "Gr00tN1d6",
3
+ "model_dtype": "bfloat16",
4
+ "model_name": "nvidia/Eagle-Block2A-2B-v2",
5
+ "backbone_model_type": "eagle",
6
+ "model_revision": null,
7
+ "tune_top_llm_layers": 4,
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": true,
15
+ "collator_overwrite_image_inputs": false,
16
+ "eagle_collator": true,
17
+ "backbone_trainable_params_fp32": true,
18
+ "extra_augmentation_config": null,
19
+ "apply_sincos_state_encoding": true,
20
+ "use_relative_action": true,
21
+ "max_state_dim": 128,
22
+ "max_action_dim": 128,
23
+ "action_horizon": 50,
24
+ "hidden_size": 1024,
25
+ "input_embedding_dim": 1536,
26
+ "add_pos_embed": true,
27
+ "attn_dropout": 0.2,
28
+ "use_vlln": true,
29
+ "max_seq_len": 1024,
30
+ "use_alternate_vl_dit": true,
31
+ "attend_text_every_n_blocks": 2,
32
+ "diffusion_model_cfg": {
33
+ "attention_head_dim": 48,
34
+ "dropout": 0.2,
35
+ "final_dropout": true,
36
+ "interleave_self_attention": true,
37
+ "norm_type": "ada_norm",
38
+ "num_attention_heads": 32,
39
+ "num_layers": 32,
40
+ "output_dim": 1024,
41
+ "positional_embeddings": null
42
+ },
43
+ "num_inference_timesteps": 4,
44
+ "noise_beta_alpha": 1.5,
45
+ "noise_beta_beta": 1.0,
46
+ "noise_s": 0.999,
47
+ "num_timestep_buckets": 1000,
48
+ "tune_projector": true,
49
+ "tune_diffusion_model": true,
50
+ "tune_vlln": true,
51
+ "state_dropout_prob": 0.0,
52
+ "state_additive_noise_scale": 0.0,
53
+ "max_num_embodiments": 32
54
+ }
checkpoint-10000/experiment_cfg/final_processor_config.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-10000/model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06070fc3c59d66e24bfd247c19112d6ee2be19fbd1af1a9bb22b70e915a0627f
3
+ size 4990571768
checkpoint-10000/model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6df75ffeb0b70c90d4c0f5a1447d27b45e85309c711358108f4f85d43ab78e8f
3
+ size 4823190320
checkpoint-10000/model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-10000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8fa1416c35d082c0e571bb6c51dea898d389db94e05dec3c786b27250dd745b
3
+ size 12960193762
checkpoint-10000/processor_config.json ADDED
@@ -0,0 +1,456 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "processor_class": "Gr00tN1d6Processor",
3
+ "processor_kwargs": {
4
+ "modality_configs": {
5
+ "behavior_r1_pro": {
6
+ "video": {
7
+ "delta_indices": [
8
+ 0
9
+ ],
10
+ "modality_keys": [
11
+ "observation.images.rgb.head_256_256",
12
+ "observation.images.rgb.left_wrist_256_256",
13
+ "observation.images.rgb.right_wrist_256_256"
14
+ ],
15
+ "sin_cos_embedding_keys": null,
16
+ "mean_std_embedding_keys": null,
17
+ "action_configs": null
18
+ },
19
+ "state": {
20
+ "delta_indices": [
21
+ 0
22
+ ],
23
+ "modality_keys": [
24
+ "robot_pos",
25
+ "robot_ori_cos",
26
+ "robot_ori_sin",
27
+ "robot_2d_ori",
28
+ "robot_2d_ori_cos",
29
+ "robot_2d_ori_sin",
30
+ "robot_lin_vel",
31
+ "robot_ang_vel",
32
+ "arm_left_qpos",
33
+ "arm_left_qpos_sin",
34
+ "arm_left_qpos_cos",
35
+ "eef_left_pos",
36
+ "eef_left_quat",
37
+ "gripper_left_qpos",
38
+ "arm_right_qpos",
39
+ "arm_right_qpos_sin",
40
+ "arm_right_qpos_cos",
41
+ "eef_right_pos",
42
+ "eef_right_quat",
43
+ "gripper_right_qpos",
44
+ "trunk_qpos"
45
+ ],
46
+ "sin_cos_embedding_keys": null,
47
+ "mean_std_embedding_keys": null,
48
+ "action_configs": null
49
+ },
50
+ "action": {
51
+ "delta_indices": [
52
+ 0,
53
+ 1,
54
+ 2,
55
+ 3,
56
+ 4,
57
+ 5,
58
+ 6,
59
+ 7,
60
+ 8,
61
+ 9,
62
+ 10,
63
+ 11,
64
+ 12,
65
+ 13,
66
+ 14,
67
+ 15,
68
+ 16,
69
+ 17,
70
+ 18,
71
+ 19,
72
+ 20,
73
+ 21,
74
+ 22,
75
+ 23,
76
+ 24,
77
+ 25,
78
+ 26,
79
+ 27,
80
+ 28,
81
+ 29,
82
+ 30,
83
+ 31
84
+ ],
85
+ "modality_keys": [
86
+ "base",
87
+ "torso",
88
+ "left_arm",
89
+ "left_gripper",
90
+ "right_arm",
91
+ "right_gripper"
92
+ ],
93
+ "sin_cos_embedding_keys": null,
94
+ "mean_std_embedding_keys": null,
95
+ "action_configs": [
96
+ {
97
+ "rep": "ABSOLUTE",
98
+ "type": "NON_EEF",
99
+ "format": "DEFAULT",
100
+ "state_key": null
101
+ },
102
+ {
103
+ "rep": "RELATIVE",
104
+ "type": "NON_EEF",
105
+ "format": "DEFAULT",
106
+ "state_key": "trunk_qpos"
107
+ },
108
+ {
109
+ "rep": "RELATIVE",
110
+ "type": "NON_EEF",
111
+ "format": "DEFAULT",
112
+ "state_key": "arm_left_qpos"
113
+ },
114
+ {
115
+ "rep": "ABSOLUTE",
116
+ "type": "NON_EEF",
117
+ "format": "DEFAULT",
118
+ "state_key": null
119
+ },
120
+ {
121
+ "rep": "RELATIVE",
122
+ "type": "NON_EEF",
123
+ "format": "DEFAULT",
124
+ "state_key": "arm_right_qpos"
125
+ },
126
+ {
127
+ "rep": "ABSOLUTE",
128
+ "type": "NON_EEF",
129
+ "format": "DEFAULT",
130
+ "state_key": null
131
+ }
132
+ ]
133
+ },
134
+ "language": {
135
+ "delta_indices": [
136
+ 0
137
+ ],
138
+ "modality_keys": [
139
+ "annotation.human.coarse_action"
140
+ ],
141
+ "sin_cos_embedding_keys": null,
142
+ "mean_std_embedding_keys": null,
143
+ "action_configs": null
144
+ }
145
+ },
146
+ "gr1": {
147
+ "video": {
148
+ "delta_indices": [
149
+ 0
150
+ ],
151
+ "modality_keys": [
152
+ "ego_view_bg_crop_pad_res256_freq20"
153
+ ],
154
+ "sin_cos_embedding_keys": null,
155
+ "mean_std_embedding_keys": null,
156
+ "action_configs": null
157
+ },
158
+ "state": {
159
+ "delta_indices": [
160
+ 0
161
+ ],
162
+ "modality_keys": [
163
+ "left_arm",
164
+ "right_arm",
165
+ "left_hand",
166
+ "right_hand",
167
+ "waist"
168
+ ],
169
+ "sin_cos_embedding_keys": [
170
+ "left_arm",
171
+ "right_arm",
172
+ "left_hand",
173
+ "right_hand",
174
+ "waist"
175
+ ],
176
+ "mean_std_embedding_keys": null,
177
+ "action_configs": null
178
+ },
179
+ "action": {
180
+ "delta_indices": [
181
+ 0,
182
+ 1,
183
+ 2,
184
+ 3,
185
+ 4,
186
+ 5,
187
+ 6,
188
+ 7,
189
+ 8,
190
+ 9,
191
+ 10,
192
+ 11,
193
+ 12,
194
+ 13,
195
+ 14,
196
+ 15
197
+ ],
198
+ "modality_keys": [
199
+ "left_arm",
200
+ "right_arm",
201
+ "left_hand",
202
+ "right_hand",
203
+ "waist"
204
+ ],
205
+ "sin_cos_embedding_keys": null,
206
+ "mean_std_embedding_keys": null,
207
+ "action_configs": [
208
+ {
209
+ "rep": "RELATIVE",
210
+ "type": "NON_EEF",
211
+ "format": "DEFAULT",
212
+ "state_key": null
213
+ },
214
+ {
215
+ "rep": "RELATIVE",
216
+ "type": "NON_EEF",
217
+ "format": "DEFAULT",
218
+ "state_key": null
219
+ },
220
+ {
221
+ "rep": "RELATIVE",
222
+ "type": "NON_EEF",
223
+ "format": "DEFAULT",
224
+ "state_key": null
225
+ },
226
+ {
227
+ "rep": "RELATIVE",
228
+ "type": "NON_EEF",
229
+ "format": "DEFAULT",
230
+ "state_key": null
231
+ },
232
+ {
233
+ "rep": "ABSOLUTE",
234
+ "type": "NON_EEF",
235
+ "format": "DEFAULT",
236
+ "state_key": null
237
+ }
238
+ ]
239
+ },
240
+ "language": {
241
+ "delta_indices": [
242
+ 0
243
+ ],
244
+ "modality_keys": [
245
+ "task"
246
+ ],
247
+ "sin_cos_embedding_keys": null,
248
+ "mean_std_embedding_keys": null,
249
+ "action_configs": null
250
+ }
251
+ },
252
+ "robocasa_panda_omron": {
253
+ "video": {
254
+ "delta_indices": [
255
+ 0
256
+ ],
257
+ "modality_keys": [
258
+ "res256_image_side_0",
259
+ "res256_image_side_1",
260
+ "res256_image_wrist_0"
261
+ ],
262
+ "sin_cos_embedding_keys": null,
263
+ "mean_std_embedding_keys": null,
264
+ "action_configs": null
265
+ },
266
+ "state": {
267
+ "delta_indices": [
268
+ 0
269
+ ],
270
+ "modality_keys": [
271
+ "end_effector_position_relative",
272
+ "end_effector_rotation_relative",
273
+ "gripper_qpos",
274
+ "base_position",
275
+ "base_rotation"
276
+ ],
277
+ "sin_cos_embedding_keys": null,
278
+ "mean_std_embedding_keys": null,
279
+ "action_configs": null
280
+ },
281
+ "action": {
282
+ "delta_indices": [
283
+ 0,
284
+ 1,
285
+ 2,
286
+ 3,
287
+ 4,
288
+ 5,
289
+ 6,
290
+ 7,
291
+ 8,
292
+ 9,
293
+ 10,
294
+ 11,
295
+ 12,
296
+ 13,
297
+ 14,
298
+ 15
299
+ ],
300
+ "modality_keys": [
301
+ "end_effector_position",
302
+ "end_effector_rotation",
303
+ "gripper_close",
304
+ "base_motion",
305
+ "control_mode"
306
+ ],
307
+ "sin_cos_embedding_keys": null,
308
+ "mean_std_embedding_keys": null,
309
+ "action_configs": [
310
+ {
311
+ "rep": "ABSOLUTE",
312
+ "type": "NON_EEF",
313
+ "format": "DEFAULT",
314
+ "state_key": null
315
+ },
316
+ {
317
+ "rep": "ABSOLUTE",
318
+ "type": "NON_EEF",
319
+ "format": "DEFAULT",
320
+ "state_key": null
321
+ },
322
+ {
323
+ "rep": "ABSOLUTE",
324
+ "type": "NON_EEF",
325
+ "format": "DEFAULT",
326
+ "state_key": null
327
+ },
328
+ {
329
+ "rep": "ABSOLUTE",
330
+ "type": "NON_EEF",
331
+ "format": "DEFAULT",
332
+ "state_key": null
333
+ },
334
+ {
335
+ "rep": "ABSOLUTE",
336
+ "type": "NON_EEF",
337
+ "format": "DEFAULT",
338
+ "state_key": null
339
+ }
340
+ ]
341
+ },
342
+ "language": {
343
+ "delta_indices": [
344
+ 0
345
+ ],
346
+ "modality_keys": [
347
+ "annotation.human.action.task_description"
348
+ ],
349
+ "sin_cos_embedding_keys": null,
350
+ "mean_std_embedding_keys": null,
351
+ "action_configs": null
352
+ }
353
+ },
354
+ "new_embodiment": {
355
+ "video": {
356
+ "delta_indices": [
357
+ 0
358
+ ],
359
+ "modality_keys": [
360
+ "front",
361
+ "wrist_color",
362
+ "wrist_depth"
363
+ ],
364
+ "sin_cos_embedding_keys": null,
365
+ "mean_std_embedding_keys": null,
366
+ "action_configs": null
367
+ },
368
+ "state": {
369
+ "delta_indices": [
370
+ 0
371
+ ],
372
+ "modality_keys": [
373
+ "single_arm",
374
+ "gripper"
375
+ ],
376
+ "sin_cos_embedding_keys": null,
377
+ "mean_std_embedding_keys": null,
378
+ "action_configs": null
379
+ },
380
+ "action": {
381
+ "delta_indices": [
382
+ 0,
383
+ 1,
384
+ 2,
385
+ 3,
386
+ 4,
387
+ 5,
388
+ 6,
389
+ 7,
390
+ 8,
391
+ 9,
392
+ 10,
393
+ 11,
394
+ 12,
395
+ 13,
396
+ 14,
397
+ 15
398
+ ],
399
+ "modality_keys": [
400
+ "single_arm",
401
+ "gripper"
402
+ ],
403
+ "sin_cos_embedding_keys": null,
404
+ "mean_std_embedding_keys": null,
405
+ "action_configs": [
406
+ {
407
+ "rep": "RELATIVE",
408
+ "type": "NON_EEF",
409
+ "format": "DEFAULT",
410
+ "state_key": null
411
+ },
412
+ {
413
+ "rep": "ABSOLUTE",
414
+ "type": "NON_EEF",
415
+ "format": "DEFAULT",
416
+ "state_key": null
417
+ }
418
+ ]
419
+ },
420
+ "language": {
421
+ "delta_indices": [
422
+ 0
423
+ ],
424
+ "modality_keys": [
425
+ "annotation.human.task_description"
426
+ ],
427
+ "sin_cos_embedding_keys": null,
428
+ "mean_std_embedding_keys": null,
429
+ "action_configs": null
430
+ }
431
+ }
432
+ },
433
+ "image_crop_size": null,
434
+ "image_target_size": null,
435
+ "use_albumentations": true,
436
+ "random_rotation_angle": null,
437
+ "color_jitter_params": {
438
+ "brightness": 0.3,
439
+ "contrast": 0.4,
440
+ "saturation": 0.5,
441
+ "hue": 0.08
442
+ },
443
+ "shortest_image_edge": 256,
444
+ "crop_fraction": 0.95,
445
+ "model_name": "nvidia/Eagle-Block2A-2B-v2",
446
+ "model_type": "eagle",
447
+ "formalize_language": true,
448
+ "max_state_dim": 128,
449
+ "max_action_dim": 128,
450
+ "max_action_horizon": 50,
451
+ "use_percentiles": false,
452
+ "clip_outliers": true,
453
+ "apply_sincos_state_encoding": true,
454
+ "use_relative_action": true
455
+ }
456
+ }
checkpoint-10000/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28c84becc673ac8b39d411bb86fc20ca979d87928663093aa2d6253d5e59eeeb
3
+ size 14645
checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d02be6d8bda4ea9c67040ed89f878acdc986bd4df3fbc60440a9d3eacca02d63
3
+ size 1465
checkpoint-10000/statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-10000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-10000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13051b66f129cdb8424a299f8db105fd14b440fde80095be1921a164894baf89
3
+ size 5713
checkpoint-10000/wandb_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"project": "finetune-gr00t-n1d6", "run_id": "output"}