viis0002 commited on
Commit
d77f03e
·
verified ·
1 Parent(s): 462d4bd

Upload folder using huggingface_hub

Browse files
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
+ }
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
+ }
experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,206 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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: null
25
+ use_albumentations_transforms: true
26
+ extra_augmentation_config: null
27
+ formalize_language: true
28
+ apply_sincos_state_encoding: false
29
+ use_relative_action: true
30
+ max_state_dim: 29
31
+ max_action_dim: 29
32
+ action_horizon: 16
33
+ hidden_size: 1024
34
+ input_embedding_dim: 1536
35
+ add_pos_embed: true
36
+ attn_dropout: 0.2
37
+ use_vlln: true
38
+ max_seq_len: 1024
39
+ use_alternate_vl_dit: true
40
+ attend_text_every_n_blocks: 2
41
+ diffusion_model_cfg:
42
+ positional_embeddings: null
43
+ num_layers: 32
44
+ num_attention_heads: 32
45
+ attention_head_dim: 48
46
+ norm_type: ada_norm
47
+ dropout: 0.2
48
+ final_dropout: true
49
+ output_dim: 1024
50
+ interleave_self_attention: true
51
+ num_inference_timesteps: 4
52
+ noise_beta_alpha: 1.5
53
+ noise_beta_beta: 1.0
54
+ noise_s: 0.999
55
+ num_timestep_buckets: 1000
56
+ tune_projector: true
57
+ tune_diffusion_model: true
58
+ tune_vlln: true
59
+ state_dropout_prob: 0.0
60
+ state_additive_noise_scale: 0.0
61
+ max_num_embodiments: 32
62
+ data:
63
+ datasets:
64
+ - dataset_paths:
65
+ - ./demo_data/ps/viis0002/cluttered_picking_bin
66
+ embodiment_tag: new_embodiment
67
+ mix_ratio: 1.0
68
+ dataset_type: physical_embodiment
69
+ val_dataset_path: null
70
+ modality_configs:
71
+ new_embodiment:
72
+ video:
73
+ delta_indices:
74
+ - 0
75
+ modality_keys:
76
+ - front
77
+ - wrist_color
78
+ sin_cos_embedding_keys: null
79
+ mean_std_embedding_keys: null
80
+ action_configs: null
81
+ state:
82
+ delta_indices:
83
+ - 0
84
+ modality_keys:
85
+ - single_arm
86
+ - gripper
87
+ sin_cos_embedding_keys: null
88
+ mean_std_embedding_keys: null
89
+ action_configs: null
90
+ action:
91
+ delta_indices:
92
+ - 0
93
+ - 1
94
+ - 2
95
+ - 3
96
+ - 4
97
+ - 5
98
+ - 6
99
+ - 7
100
+ - 8
101
+ - 9
102
+ - 10
103
+ - 11
104
+ - 12
105
+ - 13
106
+ - 14
107
+ - 15
108
+ modality_keys:
109
+ - single_arm
110
+ - gripper
111
+ sin_cos_embedding_keys: null
112
+ mean_std_embedding_keys: null
113
+ action_configs:
114
+ - rep: RELATIVE
115
+ type: NON_EEF
116
+ format: DEFAULT
117
+ state_key: null
118
+ - rep: ABSOLUTE
119
+ type: NON_EEF
120
+ format: DEFAULT
121
+ state_key: null
122
+ language:
123
+ delta_indices:
124
+ - 0
125
+ modality_keys:
126
+ - annotation.human.task_description
127
+ sin_cos_embedding_keys: null
128
+ mean_std_embedding_keys: null
129
+ action_configs: null
130
+ download_cache: false
131
+ shard_size: 1024
132
+ episode_sampling_rate: 0.1
133
+ num_shards_per_epoch: 100000
134
+ override_pretraining_statistics: false
135
+ mode: single_turn
136
+ random_chop: 0.0
137
+ mock_dataset_mode: false
138
+ shuffle: true
139
+ seed: 42
140
+ multiprocessing_context: fork
141
+ allow_padding: false
142
+ subsample_ratio: 1.0
143
+ image_crop_size:
144
+ - 244
145
+ - 244
146
+ image_target_size:
147
+ - 224
148
+ - 224
149
+ video_backend: torchcodec
150
+ training:
151
+ output_dir: ./output_RGB
152
+ experiment_name: null
153
+ max_steps: 10000
154
+ global_batch_size: 32
155
+ batch_size: null
156
+ gradient_accumulation_steps: 1
157
+ learning_rate: 0.0001
158
+ lr_scheduler_type: cosine
159
+ weight_decay: 1.0e-05
160
+ warmup_ratio: 0.05
161
+ warmup_steps: 0
162
+ max_grad_norm: 1.0
163
+ optim: adamw_torch
164
+ start_from_checkpoint: nvidia/GR00T-N1.6-3B
165
+ tf32: true
166
+ fp16: false
167
+ bf16: true
168
+ eval_bf16: true
169
+ logging_steps: 10
170
+ save_steps: 4000
171
+ save_total_limit: 5
172
+ save_vl_model: false
173
+ upload_checkpoints: false
174
+ upload_every: 1000
175
+ upload_last_n_checkpoints: 5
176
+ max_concurrent_uploads: 2
177
+ eval_strategy: 'no'
178
+ eval_steps: 500
179
+ eval_set_split_ratio: 0.1
180
+ eval_batch_size: 2
181
+ save_best_eval_metric_name: ''
182
+ save_best_eval_metric_greater_is_better: true
183
+ deepspeed_stage: 2
184
+ gradient_checkpointing: false
185
+ transformers_trust_remote_code: true
186
+ transformers_local_files_only: false
187
+ transformers_cache_dir: null
188
+ transformers_access_token: null
189
+ use_ddp: false
190
+ ddp_bucket_cap_mb: 100
191
+ num_gpus: 1
192
+ dataloader_num_workers: 4
193
+ remove_unused_columns: false
194
+ use_wandb: false
195
+ wandb_project: finetune-gr00t-n1d6
196
+ enable_profiling: false
197
+ max_retries: 3
198
+ assert_loss_less_than: null
199
+ add_rl_callback: false
200
+ enable_open_loop_eval: false
201
+ open_loop_eval_traj_ids:
202
+ - 0
203
+ open_loop_eval_steps_per_traj: 100
204
+ open_loop_eval_plot_indices: null
205
+ max_steps: 10000
206
+ save_steps: 4000
experiment_cfg/config.yaml ADDED
@@ -0,0 +1,240 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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/ps/viis0002/cluttered_picking_bin
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
+ sin_cos_embedding_keys: null
87
+ mode: single_turn
88
+ multiprocessing_context: fork
89
+ num_shards_per_epoch: 100000
90
+ override_pretraining_statistics: false
91
+ random_chop: 0.0
92
+ seed: 42
93
+ shard_size: 1024
94
+ shuffle: true
95
+ subsample_ratio: 1.0
96
+ video_backend: torchcodec
97
+ load_config_path: null
98
+ model: !!python/object:gr00t.configs.model.gr00t_n1d6.Gr00tN1d6Config
99
+ _attn_implementation_autoset: false
100
+ _attn_implementation_internal: null
101
+ _commit_hash: null
102
+ _name_or_path: ''
103
+ add_cross_attention: false
104
+ architectures: null
105
+ backbone_model_type: eagle
106
+ backbone_trainable_params_fp32: true
107
+ bad_words_ids: null
108
+ begin_suppress_tokens: null
109
+ bos_token_id: null
110
+ chunk_size_feed_forward: 0
111
+ color_jitter_params: null
112
+ cross_attention_hidden_size: null
113
+ decoder_start_token_id: null
114
+ diffusion_model_cfg:
115
+ attention_head_dim: 48
116
+ dropout: 0.2
117
+ final_dropout: true
118
+ interleave_self_attention: true
119
+ norm_type: ada_norm
120
+ num_attention_heads: 32
121
+ num_layers: 32
122
+ output_dim: 1024
123
+ positional_embeddings: null
124
+ diversity_penalty: 0.0
125
+ do_sample: false
126
+ eagle_collator: true
127
+ early_stopping: false
128
+ encoder_no_repeat_ngram_size: 0
129
+ eos_token_id: null
130
+ exponential_decay_length_penalty: null
131
+ extra_augmentation_config: null
132
+ finetuning_task: null
133
+ forced_bos_token_id: null
134
+ forced_eos_token_id: null
135
+ id2label:
136
+ 0: LABEL_0
137
+ 1: LABEL_1
138
+ is_decoder: false
139
+ is_encoder_decoder: false
140
+ label2id:
141
+ LABEL_0: 0
142
+ LABEL_1: 1
143
+ length_penalty: 1.0
144
+ load_bf16: false
145
+ max_length: 20
146
+ min_length: 0
147
+ model_name: nvidia/Eagle-Block2A-2B-v2
148
+ no_repeat_ngram_size: 0
149
+ num_beam_groups: 1
150
+ num_beams: 1
151
+ num_return_sequences: 1
152
+ output_attentions: false
153
+ output_hidden_states: false
154
+ output_scores: false
155
+ pad_token_id: null
156
+ prefix: null
157
+ problem_type: null
158
+ pruned_heads: {}
159
+ random_rotation_angle: null
160
+ remove_invalid_values: false
161
+ repetition_penalty: 1.0
162
+ reproject_vision: false
163
+ return_dict: true
164
+ return_dict_in_generate: false
165
+ sep_token_id: null
166
+ state_dropout_prob: 0.0
167
+ suppress_tokens: null
168
+ task_specific_params: null
169
+ temperature: 1.0
170
+ tf_legacy_loss: false
171
+ tie_encoder_decoder: false
172
+ tie_word_embeddings: true
173
+ tokenizer_class: null
174
+ top_k: 50
175
+ top_p: 1.0
176
+ torch_dtype: null
177
+ torchscript: false
178
+ transformers_version: null
179
+ tune_diffusion_model: true
180
+ tune_llm: false
181
+ tune_projector: true
182
+ tune_visual: false
183
+ typical_p: 1.0
184
+ use_bfloat16: false
185
+ use_relative_action: true
186
+ training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
187
+ add_rl_callback: false
188
+ assert_loss_less_than: null
189
+ batch_size: null
190
+ bf16: true
191
+ dataloader_num_workers: 4
192
+ ddp_bucket_cap_mb: 100
193
+ deepspeed_stage: 2
194
+ enable_open_loop_eval: false
195
+ enable_profiling: false
196
+ eval_batch_size: 2
197
+ eval_bf16: true
198
+ eval_set_split_ratio: 0.1
199
+ eval_steps: 500
200
+ eval_strategy: 'no'
201
+ experiment_name: null
202
+ fp16: false
203
+ global_batch_size: 32
204
+ gradient_accumulation_steps: 1
205
+ gradient_checkpointing: false
206
+ learning_rate: 0.0001
207
+ logging_steps: 10
208
+ lr_scheduler_type: cosine
209
+ max_concurrent_uploads: 2
210
+ max_grad_norm: 1.0
211
+ max_retries: 3
212
+ max_steps: 10000
213
+ num_gpus: 1
214
+ open_loop_eval_plot_indices: null
215
+ open_loop_eval_steps_per_traj: 100
216
+ open_loop_eval_traj_ids:
217
+ - 0
218
+ optim: adamw_torch
219
+ output_dir: ./output_RGB
220
+ remove_unused_columns: false
221
+ save_best_eval_metric_greater_is_better: true
222
+ save_best_eval_metric_name: ''
223
+ save_steps: 4000
224
+ save_total_limit: 5
225
+ save_vl_model: false
226
+ start_from_checkpoint: nvidia/GR00T-N1.6-3B
227
+ tf32: true
228
+ transformers_access_token: null
229
+ transformers_cache_dir: null
230
+ transformers_local_files_only: false
231
+ transformers_trust_remote_code: true
232
+ upload_checkpoints: false
233
+ upload_every: 1000
234
+ upload_last_n_checkpoints: 5
235
+ use_ddp: false
236
+ use_wandb: false
237
+ wandb_project: finetune-gr00t-n1d6
238
+ warmup_ratio: 0.05
239
+ warmup_steps: 0
240
+ weight_decay: 1.0e-05
experiment_cfg/dataset_statistics.json ADDED
@@ -0,0 +1,824 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "new_embodiment": {
3
+ "state": {
4
+ "single_arm": {
5
+ "min": [
6
+ -39.36348342895508,
7
+ -99.65782928466797,
8
+ -61.97892761230469,
9
+ 55.84988784790039,
10
+ -26.398046493530273
11
+ ],
12
+ "max": [
13
+ 7.956449031829834,
14
+ 49.1017951965332,
15
+ 100.0,
16
+ 99.55850219726562,
17
+ 57.851036071777344
18
+ ],
19
+ "mean": [
20
+ -6.385418343874992,
21
+ -24.622329268110402,
22
+ 18.56802663314117,
23
+ 81.77844942661774,
24
+ 10.85042606146218
25
+ ],
26
+ "std": [
27
+ 10.46960328595638,
28
+ 51.62350175979786,
29
+ 53.14153035933821,
30
+ 10.531352359119015,
31
+ 18.254572925078886
32
+ ],
33
+ "q01": [
34
+ -28.374080331028793,
35
+ -98.99477709610561,
36
+ -29.833800342347036,
37
+ 66.12401334198283,
38
+ -1.0385594413972399
39
+ ],
40
+ "q99": [
41
+ 1.9539289179059653,
42
+ 34.71050091958405,
43
+ 96.99906607704047,
44
+ 93.29626893615101,
45
+ 23.499264058127626
46
+ ]
47
+ },
48
+ "gripper": {
49
+ "min": [
50
+ 0.8333333134651184
51
+ ],
52
+ "max": [
53
+ 68.52564239501953
54
+ ],
55
+ "mean": [
56
+ 8.585488289392284
57
+ ],
58
+ "std": [
59
+ 11.661403495903766
60
+ ],
61
+ "q01": [
62
+ 0.966566352866377
63
+ ],
64
+ "q99": [
65
+ 42.39703828250058
66
+ ]
67
+ }
68
+ },
69
+ "action": {
70
+ "single_arm": {
71
+ "min": [
72
+ -39.73866271972656,
73
+ -100.0,
74
+ -65.7715072631836,
75
+ 54.796321868896484,
76
+ -26.886445999145508
77
+ ],
78
+ "max": [
79
+ 8.685626029968262,
80
+ 48.8510627746582,
81
+ 97.72416687011719,
82
+ 100.0,
83
+ 58.19291687011719
84
+ ],
85
+ "mean": [
86
+ -6.205533711700667,
87
+ -26.562827958046608,
88
+ 16.34602250075434,
89
+ 81.57096115074431,
90
+ 10.85811976397703
91
+ ],
92
+ "std": [
93
+ 10.59422603504957,
94
+ 50.802737184612496,
95
+ 53.54985182039301,
96
+ 10.726019282916475,
97
+ 18.349091877543735
98
+ ],
99
+ "q01": [
100
+ -28.653126963685672,
101
+ -99.79199853840126,
102
+ -32.71377093241167,
103
+ 65.14824963891643,
104
+ -1.275976052805534
105
+ ],
106
+ "q99": [
107
+ 2.4851899093151366,
108
+ 34.353396852719165,
109
+ 94.1144943032115,
110
+ 93.66894998114458,
111
+ 23.70856088634366
112
+ ]
113
+ },
114
+ "gripper": {
115
+ "min": [
116
+ 0.0
117
+ ],
118
+ "max": [
119
+ 68.81810760498047
120
+ ],
121
+ "mean": [
122
+ 6.1142540385448525
123
+ ],
124
+ "std": [
125
+ 11.503861267012631
126
+ ],
127
+ "q01": [
128
+ 0.4977736616828879
129
+ ],
130
+ "q99": [
131
+ 42.733437176746975
132
+ ]
133
+ }
134
+ },
135
+ "relative_action": {
136
+ "single_arm": {
137
+ "min": [
138
+ [
139
+ -4.869014739990234,
140
+ -15.571052551269531,
141
+ -15.518726348876953,
142
+ -7.186775207519531,
143
+ -7.228327751159668
144
+ ],
145
+ [
146
+ -6.329431533813477,
147
+ -15.571052551269531,
148
+ -18.816944122314453,
149
+ -9.114067077636719,
150
+ -8.98656940460205
151
+ ],
152
+ [
153
+ -7.716586112976074,
154
+ -16.198678970336914,
155
+ -22.023540496826172,
156
+ -10.880073547363281,
157
+ -10.647130966186523
158
+ ],
159
+ [
160
+ -9.177002906799316,
161
+ -18.16617202758789,
162
+ -25.13851547241211,
163
+ -12.646080017089844,
164
+ -12.45421028137207
165
+ ],
166
+ [
167
+ -10.517036437988281,
168
+ -20.20210075378418,
169
+ -28.233642578125,
170
+ -14.398155212402344,
171
+ -14.261293411254883
172
+ ],
173
+ [
174
+ -11.77331829071045,
175
+ -22.31884002685547,
176
+ -31.257003784179688,
177
+ -16.07585906982422,
178
+ -16.0195369720459
179
+ ],
180
+ [
181
+ -13.053546905517578,
182
+ -24.457420349121094,
183
+ -34.097137451171875,
184
+ -17.665260314941406,
185
+ -17.77777862548828
186
+ ],
187
+ [
188
+ -14.309828758239746,
189
+ -26.585079193115234,
190
+ -36.91917037963867,
191
+ -19.066925048828125,
192
+ -19.53601837158203
193
+ ],
194
+ [
195
+ -15.539650917053223,
196
+ -28.712738037109375,
197
+ -39.7411994934082,
198
+ -20.303131103515625,
199
+ -21.294261932373047
200
+ ],
201
+ [
202
+ -16.615745544433594,
203
+ -30.84039878845215,
204
+ -42.39809036254883,
205
+ -21.529586791992188,
206
+ -22.905982971191406
207
+ ],
208
+ [
209
+ -17.69184112548828,
210
+ -32.904354095458984,
211
+ -45.146602630615234,
212
+ -22.677490234375,
213
+ -24.46886444091797
214
+ ],
215
+ [
216
+ -18.696866989135742,
217
+ -35.03201675415039,
218
+ -47.786563873291016,
219
+ -23.737091064453125,
220
+ -26.178268432617188
221
+ ],
222
+ [
223
+ -19.619234085083008,
224
+ -37.085052490234375,
225
+ -50.33549499511719,
226
+ -24.613136291503906,
227
+ -27.74114990234375
228
+ ],
229
+ [
230
+ -20.541601181030273,
231
+ -39.117122650146484,
232
+ -53.05598449707031,
233
+ -25.4078369140625,
234
+ -29.206348419189453
235
+ ],
236
+ [
237
+ -21.462875366210938,
238
+ -41.17015838623047,
239
+ -55.7408561706543,
240
+ -26.10866928100586,
241
+ -30.76923179626465
242
+ ],
243
+ [
244
+ -22.30837631225586,
245
+ -43.23316955566406,
246
+ -58.580989837646484,
247
+ -26.72677230834961,
248
+ -32.23442840576172
249
+ ]
250
+ ],
251
+ "max": [
252
+ [
253
+ 6.400243759155273,
254
+ 12.026496887207031,
255
+ 10.099809646606445,
256
+ 7.688037872314453,
257
+ 9.670330047607422
258
+ ],
259
+ [
260
+ 8.159036636352539,
261
+ 15.023571014404297,
262
+ 13.306406021118164,
263
+ 8.936538696289062,
264
+ 12.014652252197266
265
+ ],
266
+ [
267
+ 9.77318000793457,
268
+ 18.217063903808594,
269
+ 16.51300048828125,
270
+ 10.86383056640625,
271
+ 14.35897445678711
272
+ ],
273
+ [
274
+ 11.425369262695312,
275
+ 21.28089141845703,
276
+ 19.627979278564453,
277
+ 12.718132019042969,
278
+ 16.75213623046875
279
+ ],
280
+ [
281
+ 13.100410461425781,
282
+ 24.27490234375,
283
+ 22.72311019897461,
284
+ 14.484138488769531,
285
+ 19.047618865966797
286
+ ],
287
+ [
288
+ 14.714553833007812,
289
+ 27.253623962402344,
290
+ 25.818239212036133,
291
+ 16.236221313476562,
292
+ 21.24542236328125
293
+ ],
294
+ [
295
+ 16.222091674804688,
296
+ 30.147239685058594,
297
+ 28.913368225097656,
298
+ 17.913925170898438,
299
+ 23.394384384155273
300
+ ],
301
+ [
302
+ 17.682506561279297,
303
+ 33.04085922241211,
304
+ 32.00849914550781,
305
+ 19.40319061279297,
306
+ 25.738704681396484
307
+ ],
308
+ [
309
+ 19.106292724609375,
310
+ 35.849369049072266,
311
+ 35.031864166259766,
312
+ 20.727691650390625,
313
+ 28.083026885986328
314
+ ],
315
+ [
316
+ 20.446325302124023,
317
+ 38.5012092590332,
318
+ 38.05522537231445,
319
+ 21.875595092773438,
320
+ 30.28083038330078
321
+ ],
322
+ [
323
+ 21.753013610839844,
324
+ 41.277400970458984,
325
+ 41.059322357177734,
326
+ 23.014450073242188,
327
+ 32.42979431152344
328
+ ],
329
+ [
330
+ 23.05970001220703,
331
+ 43.92923355102539,
332
+ 43.97238540649414,
333
+ 24.074050903320312,
334
+ 34.57875442504883
335
+ ],
336
+ [
337
+ 24.232229232788086,
338
+ 46.58107376098633,
339
+ 46.97648239135742,
340
+ 25.12529754638672,
341
+ 36.43467712402344
342
+ ],
343
+ [
344
+ 25.321006774902344,
345
+ 49.2193717956543,
346
+ 49.88954544067383,
347
+ 26.00133514404297,
348
+ 38.43711853027344
349
+ ],
350
+ [
351
+ 26.32023811340332,
352
+ 51.857669830322266,
353
+ 52.802608489990234,
354
+ 26.807876586914062,
355
+ 40.29304122924805
356
+ ],
357
+ [
358
+ 27.319469451904297,
359
+ 54.32575225830078,
360
+ 55.806705474853516,
361
+ 27.771522521972656,
362
+ 41.904762268066406
363
+ ]
364
+ ],
365
+ "mean": [
366
+ [
367
+ 0.1785251498222351,
368
+ -1.9557442665100098,
369
+ -2.2593743801116943,
370
+ -0.20868544280529022,
371
+ 0.008650688454508781
372
+ ],
373
+ [
374
+ 0.1787870079278946,
375
+ -1.9535331726074219,
376
+ -2.2614686489105225,
377
+ -0.20898747444152832,
378
+ 0.008710959926247597
379
+ ],
380
+ [
381
+ 0.17905102670192719,
382
+ -1.9514031410217285,
383
+ -2.2634711265563965,
384
+ -0.20927496254444122,
385
+ 0.008769187144935131
386
+ ],
387
+ [
388
+ 0.17931483685970306,
389
+ -1.9493659734725952,
390
+ -2.265377998352051,
391
+ -0.20954172313213348,
392
+ 0.008825326338410378
393
+ ],
394
+ [
395
+ 0.1795794814825058,
396
+ -1.947402000427246,
397
+ -2.26723575592041,
398
+ -0.20978474617004395,
399
+ 0.008879491128027439
400
+ ],
401
+ [
402
+ 0.17984697222709656,
403
+ -1.945528268814087,
404
+ -2.2690117359161377,
405
+ -0.20999972522258759,
406
+ 0.008933082222938538
407
+ ],
408
+ [
409
+ 0.18011674284934998,
410
+ -1.9437333345413208,
411
+ -2.2707107067108154,
412
+ -0.2101903259754181,
413
+ 0.008988128043711185
414
+ ],
415
+ [
416
+ 0.18038682639598846,
417
+ -1.942039132118225,
418
+ -2.2723309993743896,
419
+ -0.21035927534103394,
420
+ 0.009043456986546516
421
+ ],
422
+ [
423
+ 0.18065497279167175,
424
+ -1.9404367208480835,
425
+ -2.2738921642303467,
426
+ -0.21050825715065002,
427
+ 0.009101650677621365
428
+ ],
429
+ [
430
+ 0.18092294037342072,
431
+ -1.9389110803604126,
432
+ -2.275402545928955,
433
+ -0.21062998473644257,
434
+ 0.009162607602775097
435
+ ],
436
+ [
437
+ 0.18119066953659058,
438
+ -1.937473177909851,
439
+ -2.2768537998199463,
440
+ -0.210730642080307,
441
+ 0.009221411310136318
442
+ ],
443
+ [
444
+ 0.18145936727523804,
445
+ -1.9361056089401245,
446
+ -2.2782578468322754,
447
+ -0.21082045137882233,
448
+ 0.009277415461838245
449
+ ],
450
+ [
451
+ 0.18173053860664368,
452
+ -1.9348115921020508,
453
+ -2.2796051502227783,
454
+ -0.21090328693389893,
455
+ 0.00933289248496294
456
+ ],
457
+ [
458
+ 0.18200388550758362,
459
+ -1.9335931539535522,
460
+ -2.280888795852661,
461
+ -0.21098707616329193,
462
+ 0.009389038197696209
463
+ ],
464
+ [
465
+ 0.18227829039096832,
466
+ -1.9324415922164917,
467
+ -2.2821178436279297,
468
+ -0.2110622674226761,
469
+ 0.009444031864404678
470
+ ],
471
+ [
472
+ 0.18255291879177094,
473
+ -1.9313602447509766,
474
+ -2.2832837104797363,
475
+ -0.21112234890460968,
476
+ 0.009497801773250103
477
+ ]
478
+ ],
479
+ "std": [
480
+ [
481
+ 0.9654352068901062,
482
+ 3.4649360179901123,
483
+ 3.150923252105713,
484
+ 1.1157993078231812,
485
+ 1.0172107219696045
486
+ ],
487
+ [
488
+ 1.1931160688400269,
489
+ 4.058739185333252,
490
+ 3.8396174907684326,
491
+ 1.3442760705947876,
492
+ 1.2842698097229004
493
+ ],
494
+ [
495
+ 1.4228914976119995,
496
+ 4.675368785858154,
497
+ 4.540013790130615,
498
+ 1.574043869972229,
499
+ 1.5493543148040771
500
+ ],
501
+ [
502
+ 1.6525599956512451,
503
+ 5.304614543914795,
504
+ 5.245314598083496,
505
+ 1.8014473915100098,
506
+ 1.8111250400543213
507
+ ],
508
+ [
509
+ 1.8808844089508057,
510
+ 5.940498352050781,
511
+ 5.951900005340576,
512
+ 2.0248169898986816,
513
+ 2.0692081451416016
514
+ ],
515
+ [
516
+ 2.1071860790252686,
517
+ 6.579427242279053,
518
+ 6.656661510467529,
519
+ 2.243411064147949,
520
+ 2.323486804962158
521
+ ],
522
+ [
523
+ 2.331037759780884,
524
+ 7.219012260437012,
525
+ 7.358837604522705,
526
+ 2.4567863941192627,
527
+ 2.5739219188690186
528
+ ],
529
+ [
530
+ 2.552251100540161,
531
+ 7.857221603393555,
532
+ 8.057107925415039,
533
+ 2.664818286895752,
534
+ 2.8202738761901855
535
+ ],
536
+ [
537
+ 2.7706377506256104,
538
+ 8.493036270141602,
539
+ 8.75144100189209,
540
+ 2.8674299716949463,
541
+ 3.0627496242523193
542
+ ],
543
+ [
544
+ 2.986109733581543,
545
+ 9.125910758972168,
546
+ 9.44082260131836,
547
+ 3.064739465713501,
548
+ 3.301175355911255
549
+ ],
550
+ [
551
+ 3.198592185974121,
552
+ 9.754792213439941,
553
+ 10.125152587890625,
554
+ 3.2569146156311035,
555
+ 3.535595178604126
556
+ ],
557
+ [
558
+ 3.408093214035034,
559
+ 10.379477500915527,
560
+ 10.80418872833252,
561
+ 3.443956136703491,
562
+ 3.7658936977386475
563
+ ],
564
+ [
565
+ 3.614506959915161,
566
+ 10.999598503112793,
567
+ 11.477699279785156,
568
+ 3.6260931491851807,
569
+ 3.992025852203369
570
+ ],
571
+ [
572
+ 3.8178415298461914,
573
+ 11.614961624145508,
574
+ 12.145866394042969,
575
+ 3.8034627437591553,
576
+ 4.214049816131592
577
+ ],
578
+ [
579
+ 4.01806116104126,
580
+ 12.225295066833496,
581
+ 12.80855655670166,
582
+ 3.9761598110198975,
583
+ 4.431927680969238
584
+ ],
585
+ [
586
+ 4.215169906616211,
587
+ 12.830470085144043,
588
+ 13.465753555297852,
589
+ 4.144509792327881,
590
+ 4.6458001136779785
591
+ ]
592
+ ],
593
+ "q01": [
594
+ [
595
+ -3.1989338397979736,
596
+ -9.316668510437012,
597
+ -11.341864585876465,
598
+ -3.8126816749572754,
599
+ -3.5653228759765625
600
+ ],
601
+ [
602
+ -4.159060478210449,
603
+ -10.740071296691895,
604
+ -13.675247192382812,
605
+ -4.741844177246094,
606
+ -4.590963363647461
607
+ ],
608
+ [
609
+ -5.090120315551758,
610
+ -12.214762687683105,
611
+ -15.99472713470459,
612
+ -5.634047508239746,
613
+ -5.612699031829834
614
+ ],
615
+ [
616
+ -6.020403861999512,
617
+ -13.759285926818848,
618
+ -18.29571533203125,
619
+ -6.501101493835449,
620
+ -6.593406677246094
621
+ ],
622
+ [
623
+ -6.948093891143799,
624
+ -15.280621528625488,
625
+ -20.582651138305664,
626
+ -7.327765464782715,
627
+ -7.5702056884765625
628
+ ],
629
+ [
630
+ -7.856415271759033,
631
+ -16.814342498779297,
632
+ -22.871421813964844,
633
+ -8.132277488708496,
634
+ -8.4981689453125
635
+ ],
636
+ [
637
+ -8.780505180358887,
638
+ -18.380815505981445,
639
+ -25.133073806762695,
640
+ -8.922734260559082,
641
+ -9.474967956542969
642
+ ],
643
+ [
644
+ -9.661839485168457,
645
+ -19.964109420776367,
646
+ -27.354209899902344,
647
+ -9.681880950927734,
648
+ -10.35409164428711
649
+ ],
650
+ [
651
+ -10.548376083374023,
652
+ -21.490734100341797,
653
+ -29.52994728088379,
654
+ -10.417035102844238,
655
+ -11.233211517333984
656
+ ],
657
+ [
658
+ -11.400496482849121,
659
+ -23.03583335876465,
660
+ -31.718549728393555,
661
+ -11.105175971984863,
662
+ -12.161170959472656
663
+ ],
664
+ [
665
+ -12.23694896697998,
666
+ -24.55828094482422,
667
+ -33.89005661010742,
668
+ -11.810083389282227,
669
+ -13.085226058959961
670
+ ],
671
+ [
672
+ -13.08798599243164,
673
+ -26.0482120513916,
674
+ -36.02480697631836,
675
+ -12.4576416015625,
676
+ -13.968254089355469
677
+ ],
678
+ [
679
+ -13.868152618408203,
680
+ -27.571319580078125,
681
+ -38.08620071411133,
682
+ -13.088216781616211,
683
+ -14.847375869750977
684
+ ],
685
+ [
686
+ -14.652140617370605,
687
+ -29.0412540435791,
688
+ -40.186424255371094,
689
+ -13.666051864624023,
690
+ -15.775335311889648
691
+ ],
692
+ [
693
+ -15.41779899597168,
694
+ -30.49142837524414,
695
+ -42.280601501464844,
696
+ -14.259786605834961,
697
+ -16.60561752319336
698
+ ],
699
+ [
700
+ -16.16594886779785,
701
+ -31.97154426574707,
702
+ -44.370018005371094,
703
+ -14.746328353881836,
704
+ -17.484739303588867
705
+ ]
706
+ ],
707
+ "q99": [
708
+ [
709
+ 3.2479984760284424,
710
+ 7.912990093231201,
711
+ 5.633664608001709,
712
+ 3.4832823276519775,
713
+ 5.372402191162109
714
+ ],
715
+ [
716
+ 4.097276210784912,
717
+ 10.157513618469238,
718
+ 7.765993118286133,
719
+ 4.481320381164551,
720
+ 6.837606430053711
721
+ ],
722
+ [
723
+ 4.946655750274658,
724
+ 12.368200302124023,
725
+ 9.88430404663086,
726
+ 5.5072197914123535,
727
+ 8.298991203308105
728
+ ],
729
+ [
730
+ 5.767092704772949,
731
+ 14.587282180786133,
732
+ 11.932137489318848,
733
+ 6.434570789337158,
734
+ 9.715353965759277
735
+ ],
736
+ [
737
+ 6.5880303382873535,
738
+ 16.77672004699707,
739
+ 14.06669807434082,
740
+ 7.37797737121582,
741
+ 11.037851333618164
742
+ ],
743
+ [
744
+ 7.4164137840271,
745
+ 18.925695419311523,
746
+ 16.074583053588867,
747
+ 8.235647201538086,
748
+ 12.454212188720703
749
+ ],
750
+ [
751
+ 8.199186325073242,
752
+ 21.071264266967773,
753
+ 18.116865158081055,
754
+ 9.053117752075195,
755
+ 13.724054336547852
756
+ ],
757
+ [
758
+ 8.967052459716797,
759
+ 23.139829635620117,
760
+ 20.07178497314453,
761
+ 9.901945114135742,
762
+ 15.087759971618652
763
+ ],
764
+ [
765
+ 9.712544441223145,
766
+ 25.245664596557617,
767
+ 22.007034301757812,
768
+ 10.643431663513184,
769
+ 16.21489715576172
770
+ ],
771
+ [
772
+ 10.44395923614502,
773
+ 27.225433349609375,
774
+ 23.908384323120117,
775
+ 11.467199325561523,
776
+ 17.484737396240234
777
+ ],
778
+ [
779
+ 11.168615341186523,
780
+ 29.31740951538086,
781
+ 25.754091262817383,
782
+ 12.12663745880127,
783
+ 18.65690040588379
784
+ ],
785
+ [
786
+ 11.87579345703125,
787
+ 31.247526168823242,
788
+ 27.57501983642578,
789
+ 12.88122844696045,
790
+ 19.776405334472656
791
+ ],
792
+ [
793
+ 12.570459365844727,
794
+ 33.20111846923828,
795
+ 29.390729904174805,
796
+ 13.581562042236328,
797
+ 20.854700088500977
798
+ ],
799
+ [
800
+ 13.26676082611084,
801
+ 35.10786819458008,
802
+ 31.10728645324707,
803
+ 14.265336036682129,
804
+ 21.733821868896484
805
+ ],
806
+ [
807
+ 13.936410903930664,
808
+ 36.96277618408203,
809
+ 32.90301513671875,
810
+ 14.87486743927002,
811
+ 22.905981063842773
812
+ ],
813
+ [
814
+ 14.615703582763672,
815
+ 38.77449417114258,
816
+ 34.64031219482422,
817
+ 15.497631072998047,
818
+ 23.78510284423828
819
+ ]
820
+ ]
821
+ }
822
+ }
823
+ }
824
+ }
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
+ }
experiment_cfg/final_processor_config.json ADDED
The diff for this file is too large to render. See raw diff
 
model-00001-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f657b16c32e74a5caa3f1b985d0466c1dd06b7ed39adcfbb26c9d967975859c5
3
+ size 4990120184
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e91421ab3c69ef8eee1dbfc65b30a3143f24abea48061c1cb7a75abd4bcf82a8
3
+ size 4823190320
model.safetensors.index.json ADDED
The diff for this file is too large to render. See raw diff
 
optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2976e4c31a933c757f52bd6117ccc388d4855e22a11929d7831a1e1aea76c310
3
+ size 12960193762
processor_config.json ADDED
@@ -0,0 +1,455 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ ],
363
+ "sin_cos_embedding_keys": null,
364
+ "mean_std_embedding_keys": null,
365
+ "action_configs": null
366
+ },
367
+ "state": {
368
+ "delta_indices": [
369
+ 0
370
+ ],
371
+ "modality_keys": [
372
+ "single_arm",
373
+ "gripper"
374
+ ],
375
+ "sin_cos_embedding_keys": null,
376
+ "mean_std_embedding_keys": null,
377
+ "action_configs": null
378
+ },
379
+ "action": {
380
+ "delta_indices": [
381
+ 0,
382
+ 1,
383
+ 2,
384
+ 3,
385
+ 4,
386
+ 5,
387
+ 6,
388
+ 7,
389
+ 8,
390
+ 9,
391
+ 10,
392
+ 11,
393
+ 12,
394
+ 13,
395
+ 14,
396
+ 15
397
+ ],
398
+ "modality_keys": [
399
+ "single_arm",
400
+ "gripper"
401
+ ],
402
+ "sin_cos_embedding_keys": null,
403
+ "mean_std_embedding_keys": null,
404
+ "action_configs": [
405
+ {
406
+ "rep": "RELATIVE",
407
+ "type": "NON_EEF",
408
+ "format": "DEFAULT",
409
+ "state_key": null
410
+ },
411
+ {
412
+ "rep": "ABSOLUTE",
413
+ "type": "NON_EEF",
414
+ "format": "DEFAULT",
415
+ "state_key": null
416
+ }
417
+ ]
418
+ },
419
+ "language": {
420
+ "delta_indices": [
421
+ 0
422
+ ],
423
+ "modality_keys": [
424
+ "annotation.human.task_description"
425
+ ],
426
+ "sin_cos_embedding_keys": null,
427
+ "mean_std_embedding_keys": null,
428
+ "action_configs": null
429
+ }
430
+ }
431
+ },
432
+ "image_crop_size": null,
433
+ "image_target_size": null,
434
+ "use_albumentations": true,
435
+ "random_rotation_angle": null,
436
+ "color_jitter_params": {
437
+ "brightness": 0.3,
438
+ "contrast": 0.4,
439
+ "saturation": 0.5,
440
+ "hue": 0.08
441
+ },
442
+ "shortest_image_edge": 256,
443
+ "crop_fraction": 0.95,
444
+ "model_name": "nvidia/Eagle-Block2A-2B-v2",
445
+ "model_type": "eagle",
446
+ "formalize_language": true,
447
+ "max_state_dim": 128,
448
+ "max_action_dim": 128,
449
+ "max_action_horizon": 50,
450
+ "use_percentiles": false,
451
+ "clip_outliers": true,
452
+ "apply_sincos_state_encoding": true,
453
+ "use_relative_action": true
454
+ }
455
+ }
rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:20a0c565bd08e39a00dedd5287e1c398c93d4b9bfb3068b49a86d6103c37e0a0
3
+ size 14645
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de7de3f59b6396d7d4468481ee0f0884cb3ddcb21a44cf41b6fd5a4c73ee0149
3
+ size 1465
statistics.json ADDED
The diff for this file is too large to render. See raw diff
 
trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8ccdc4abdd7c9589099fa367b7addfc5f3480910fc3ed6e9c5a6f87301789adc
3
+ size 5713
wandb_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"project": "finetune-gr00t-n1d6", "run_id": "output_RGB"}