liorbenhorin-nv commited on
Commit
3427762
·
verified ·
1 Parent(s): 25d291c

GR00T-N1.6 fine-tune on SO-101 vials_rack_left sim+real (10k steps)

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 +1,10 @@
 
 
 
 
 
 
 
 
 
 
 
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
+ "new_embodiment": 10
10
+ }
experiment_cfg/conf.yaml ADDED
@@ -0,0 +1,209 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ formalize_language: true
31
+ apply_sincos_state_encoding: false
32
+ use_relative_action: true
33
+ max_state_dim: 29
34
+ max_action_dim: 29
35
+ action_horizon: 16
36
+ hidden_size: 1024
37
+ input_embedding_dim: 1536
38
+ add_pos_embed: true
39
+ attn_dropout: 0.2
40
+ use_vlln: true
41
+ max_seq_len: 1024
42
+ use_alternate_vl_dit: true
43
+ attend_text_every_n_blocks: 2
44
+ diffusion_model_cfg:
45
+ positional_embeddings: null
46
+ num_layers: 32
47
+ num_attention_heads: 32
48
+ attention_head_dim: 48
49
+ norm_type: ada_norm
50
+ dropout: 0.2
51
+ final_dropout: true
52
+ output_dim: 1024
53
+ interleave_self_attention: true
54
+ num_inference_timesteps: 4
55
+ noise_beta_alpha: 1.5
56
+ noise_beta_beta: 1.0
57
+ noise_s: 0.999
58
+ num_timestep_buckets: 1000
59
+ tune_projector: true
60
+ tune_diffusion_model: true
61
+ tune_vlln: true
62
+ state_dropout_prob: 0.0
63
+ state_additive_noise_scale: 0.0
64
+ max_num_embodiments: 32
65
+ data:
66
+ datasets:
67
+ - dataset_paths:
68
+ - examples/SO100/sreetz-nv/so101_teleop_vials_rack_left_sim_and_real
69
+ embodiment_tag: new_embodiment
70
+ mix_ratio: 1.0
71
+ dataset_type: physical_embodiment
72
+ val_dataset_path: null
73
+ modality_configs:
74
+ new_embodiment:
75
+ video:
76
+ delta_indices:
77
+ - 0
78
+ modality_keys:
79
+ - front
80
+ - wrist
81
+ sin_cos_embedding_keys: null
82
+ mean_std_embedding_keys: null
83
+ action_configs: null
84
+ state:
85
+ delta_indices:
86
+ - 0
87
+ modality_keys:
88
+ - single_arm
89
+ - gripper
90
+ sin_cos_embedding_keys: null
91
+ mean_std_embedding_keys: null
92
+ action_configs: null
93
+ action:
94
+ delta_indices:
95
+ - 0
96
+ - 1
97
+ - 2
98
+ - 3
99
+ - 4
100
+ - 5
101
+ - 6
102
+ - 7
103
+ - 8
104
+ - 9
105
+ - 10
106
+ - 11
107
+ - 12
108
+ - 13
109
+ - 14
110
+ - 15
111
+ modality_keys:
112
+ - single_arm
113
+ - gripper
114
+ sin_cos_embedding_keys: null
115
+ mean_std_embedding_keys: null
116
+ action_configs:
117
+ - rep: RELATIVE
118
+ type: NON_EEF
119
+ format: DEFAULT
120
+ state_key: null
121
+ - rep: ABSOLUTE
122
+ type: NON_EEF
123
+ format: DEFAULT
124
+ state_key: null
125
+ language:
126
+ delta_indices:
127
+ - 0
128
+ modality_keys:
129
+ - annotation.human.task_description
130
+ sin_cos_embedding_keys: null
131
+ mean_std_embedding_keys: null
132
+ action_configs: null
133
+ download_cache: false
134
+ shard_size: 1024
135
+ episode_sampling_rate: 0.1
136
+ num_shards_per_epoch: 100000
137
+ override_pretraining_statistics: false
138
+ mode: single_turn
139
+ random_chop: 0.0
140
+ mock_dataset_mode: false
141
+ shuffle: true
142
+ seed: 42
143
+ multiprocessing_context: fork
144
+ allow_padding: false
145
+ subsample_ratio: 1.0
146
+ image_crop_size:
147
+ - 244
148
+ - 244
149
+ image_target_size:
150
+ - 224
151
+ - 224
152
+ video_backend: torchcodec
153
+ training:
154
+ output_dir: /ephemeral/grootn16-finetune_sreetz-so101_teleop_vials_rack_left_sim_and_real
155
+ experiment_name: null
156
+ max_steps: 10000
157
+ global_batch_size: 32
158
+ batch_size: null
159
+ gradient_accumulation_steps: 1
160
+ learning_rate: 0.0001
161
+ lr_scheduler_type: cosine
162
+ weight_decay: 1.0e-05
163
+ warmup_ratio: 0.05
164
+ warmup_steps: 0
165
+ max_grad_norm: 1.0
166
+ optim: adamw_torch
167
+ start_from_checkpoint: nvidia/GR00T-N1.6-3B
168
+ tf32: true
169
+ fp16: false
170
+ bf16: true
171
+ eval_bf16: true
172
+ logging_steps: 10
173
+ save_steps: 5000
174
+ save_total_limit: 5
175
+ save_vl_model: false
176
+ upload_checkpoints: false
177
+ upload_every: 1000
178
+ upload_last_n_checkpoints: 5
179
+ max_concurrent_uploads: 2
180
+ eval_strategy: 'no'
181
+ eval_steps: 500
182
+ eval_set_split_ratio: 0.1
183
+ eval_batch_size: 2
184
+ save_best_eval_metric_name: ''
185
+ save_best_eval_metric_greater_is_better: true
186
+ deepspeed_stage: 2
187
+ gradient_checkpointing: false
188
+ transformers_trust_remote_code: true
189
+ transformers_local_files_only: false
190
+ transformers_cache_dir: null
191
+ transformers_access_token: null
192
+ use_ddp: false
193
+ ddp_bucket_cap_mb: 100
194
+ num_gpus: 1
195
+ dataloader_num_workers: 4
196
+ remove_unused_columns: false
197
+ use_wandb: false
198
+ wandb_project: finetune-gr00t-n1d6
199
+ enable_profiling: false
200
+ max_retries: 3
201
+ assert_loss_less_than: null
202
+ add_rl_callback: false
203
+ enable_open_loop_eval: false
204
+ open_loop_eval_traj_ids:
205
+ - 0
206
+ open_loop_eval_steps_per_traj: 100
207
+ open_loop_eval_plot_indices: null
208
+ max_steps: 10000
209
+ save_steps: 5000
experiment_cfg/config.yaml ADDED
@@ -0,0 +1,243 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ - examples/SO100/sreetz-nv/so101_teleop_vials_rack_left_sim_and_real
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
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:
112
+ brightness: 0.3
113
+ contrast: 0.4
114
+ hue: 0.08
115
+ saturation: 0.5
116
+ cross_attention_hidden_size: null
117
+ decoder_start_token_id: null
118
+ diffusion_model_cfg:
119
+ attention_head_dim: 48
120
+ dropout: 0.2
121
+ final_dropout: true
122
+ interleave_self_attention: true
123
+ norm_type: ada_norm
124
+ num_attention_heads: 32
125
+ num_layers: 32
126
+ output_dim: 1024
127
+ positional_embeddings: null
128
+ diversity_penalty: 0.0
129
+ do_sample: false
130
+ eagle_collator: true
131
+ early_stopping: false
132
+ encoder_no_repeat_ngram_size: 0
133
+ eos_token_id: null
134
+ exponential_decay_length_penalty: null
135
+ finetuning_task: null
136
+ forced_bos_token_id: null
137
+ forced_eos_token_id: null
138
+ id2label:
139
+ 0: LABEL_0
140
+ 1: LABEL_1
141
+ is_decoder: false
142
+ is_encoder_decoder: false
143
+ label2id:
144
+ LABEL_0: 0
145
+ LABEL_1: 1
146
+ length_penalty: 1.0
147
+ load_bf16: false
148
+ max_length: 20
149
+ min_length: 0
150
+ model_name: nvidia/Eagle-Block2A-2B-v2
151
+ no_repeat_ngram_size: 0
152
+ num_beam_groups: 1
153
+ num_beams: 1
154
+ num_return_sequences: 1
155
+ output_attentions: false
156
+ output_hidden_states: false
157
+ output_scores: false
158
+ pad_token_id: null
159
+ prefix: null
160
+ problem_type: null
161
+ pruned_heads: {}
162
+ random_rotation_angle: null
163
+ remove_invalid_values: false
164
+ repetition_penalty: 1.0
165
+ reproject_vision: false
166
+ return_dict: true
167
+ return_dict_in_generate: false
168
+ sep_token_id: null
169
+ state_dropout_prob: 0.0
170
+ suppress_tokens: null
171
+ task_specific_params: null
172
+ temperature: 1.0
173
+ tf_legacy_loss: false
174
+ tie_encoder_decoder: false
175
+ tie_word_embeddings: true
176
+ tokenizer_class: null
177
+ top_k: 50
178
+ top_p: 1.0
179
+ torch_dtype: null
180
+ torchscript: false
181
+ transformers_version: null
182
+ tune_diffusion_model: true
183
+ tune_llm: false
184
+ tune_projector: true
185
+ tune_visual: false
186
+ typical_p: 1.0
187
+ use_bfloat16: false
188
+ use_relative_action: true
189
+ training: !!python/object:gr00t.configs.training.training_config.TrainingConfig
190
+ add_rl_callback: false
191
+ assert_loss_less_than: null
192
+ batch_size: null
193
+ bf16: true
194
+ dataloader_num_workers: 4
195
+ ddp_bucket_cap_mb: 100
196
+ deepspeed_stage: 2
197
+ enable_open_loop_eval: false
198
+ enable_profiling: false
199
+ eval_batch_size: 2
200
+ eval_bf16: true
201
+ eval_set_split_ratio: 0.1
202
+ eval_steps: 500
203
+ eval_strategy: 'no'
204
+ experiment_name: null
205
+ fp16: false
206
+ global_batch_size: 32
207
+ gradient_accumulation_steps: 1
208
+ gradient_checkpointing: false
209
+ learning_rate: 0.0001
210
+ logging_steps: 10
211
+ lr_scheduler_type: cosine
212
+ max_concurrent_uploads: 2
213
+ max_grad_norm: 1.0
214
+ max_retries: 3
215
+ max_steps: 10000
216
+ num_gpus: 1
217
+ open_loop_eval_plot_indices: null
218
+ open_loop_eval_steps_per_traj: 100
219
+ open_loop_eval_traj_ids:
220
+ - 0
221
+ optim: adamw_torch
222
+ output_dir: /ephemeral/grootn16-finetune_sreetz-so101_teleop_vials_rack_left_sim_and_real
223
+ remove_unused_columns: false
224
+ save_best_eval_metric_greater_is_better: true
225
+ save_best_eval_metric_name: ''
226
+ save_steps: 5000
227
+ save_total_limit: 5
228
+ save_vl_model: false
229
+ start_from_checkpoint: nvidia/GR00T-N1.6-3B
230
+ tf32: true
231
+ transformers_access_token: null
232
+ transformers_cache_dir: null
233
+ transformers_local_files_only: false
234
+ transformers_trust_remote_code: true
235
+ upload_checkpoints: false
236
+ upload_every: 1000
237
+ upload_last_n_checkpoints: 5
238
+ use_ddp: false
239
+ use_wandb: false
240
+ wandb_project: finetune-gr00t-n1d6
241
+ warmup_ratio: 0.05
242
+ warmup_steps: 0
243
+ 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
+ -44.03818893432617,
7
+ -98.77899932861328,
8
+ -34.79449462890625,
9
+ -100.0046157836914,
10
+ -71.68205261230469
11
+ ],
12
+ "max": [
13
+ 35.02424621582031,
14
+ 50.342254638671875,
15
+ 100.00312805175781,
16
+ 66.6961669921875,
17
+ -34.87779235839844
18
+ ],
19
+ "mean": [
20
+ -5.529948090689649,
21
+ -32.25617507398843,
22
+ 59.49486274156851,
23
+ -3.685691143103243,
24
+ -54.93849806281332
25
+ ],
26
+ "std": [
27
+ 19.169057558994496,
28
+ 36.501744311889304,
29
+ 27.44182687675344,
30
+ 44.11783042382442,
31
+ 5.004174026389249
32
+ ],
33
+ "q01": [
34
+ -28.133388865368303,
35
+ -94.34519707333364,
36
+ 15.038800355953908,
37
+ -72.69508678415208,
38
+ -59.0400376459965
39
+ ],
40
+ "q99": [
41
+ 16.99107606037438,
42
+ 21.652577595572843,
43
+ 90.42731385148896,
44
+ 41.45560175824795,
45
+ -48.75060160793834
46
+ ]
47
+ },
48
+ "gripper": {
49
+ "min": [
50
+ 8.497326850891113
51
+ ],
52
+ "max": [
53
+ 58.09996795654297
54
+ ],
55
+ "mean": [
56
+ 25.875692358407623
57
+ ],
58
+ "std": [
59
+ 9.941198851120081
60
+ ],
61
+ "q01": [
62
+ 16.605927441558602
63
+ ],
64
+ "q99": [
65
+ 41.45135965913047
66
+ ]
67
+ }
68
+ },
69
+ "action": {
70
+ "single_arm": {
71
+ "min": [
72
+ -44.09799575805664,
73
+ -99.24528503417969,
74
+ -35.0520133972168,
75
+ -99.82714080810547,
76
+ -72.86458587646484
77
+ ],
78
+ "max": [
79
+ 34.966590881347656,
80
+ 49.93710708618164,
81
+ 99.9095458984375,
82
+ 67.07000732421875,
83
+ -34.84375
84
+ ],
85
+ "mean": [
86
+ -5.659139500292856,
87
+ -32.84102569102923,
88
+ 58.508223843381145,
89
+ -4.174387269112394,
90
+ -54.96607035060066
91
+ ],
92
+ "std": [
93
+ 19.208448958208123,
94
+ 36.21666989550831,
95
+ 27.483782442614924,
96
+ 44.269929238817085,
97
+ 5.102061272612929
98
+ ],
99
+ "q01": [
100
+ -28.216125322018698,
101
+ -94.58987980673707,
102
+ 14.437293018114238,
103
+ -72.98535552267427,
104
+ -59.132343563414366
105
+ ],
106
+ "q99": [
107
+ 16.910335114307024,
108
+ 21.40291647358651,
109
+ 89.70153425583858,
110
+ 41.51907899051743,
111
+ -48.575680159378805
112
+ ]
113
+ },
114
+ "gripper": {
115
+ "min": [
116
+ 0.4850444495677948
117
+ ],
118
+ "max": [
119
+ 58.69038009643555
120
+ ],
121
+ "mean": [
122
+ 20.726677986393724
123
+ ],
124
+ "std": [
125
+ 15.457074618810392
126
+ ],
127
+ "q01": [
128
+ 4.2216537105235385
129
+ ],
130
+ "q99": [
131
+ 41.74233933712091
132
+ ]
133
+ }
134
+ },
135
+ "relative_action": {
136
+ "single_arm": {
137
+ "min": [
138
+ [
139
+ -7.403679847717285,
140
+ -21.343788146972656,
141
+ -11.686344146728516,
142
+ -26.58721160888672,
143
+ -4.000480651855469
144
+ ],
145
+ [
146
+ -15.338791847229004,
147
+ -30.557064056396484,
148
+ -13.969884872436523,
149
+ -36.63103485107422,
150
+ -5.058338165283203
151
+ ],
152
+ [
153
+ -17.60955810546875,
154
+ -33.57593536376953,
155
+ -16.110721588134766,
156
+ -39.94593811035156,
157
+ -6.100704193115234
158
+ ],
159
+ [
160
+ -19.56610107421875,
161
+ -36.97331619262695,
162
+ -18.439834594726562,
163
+ -43.18556213378906,
164
+ -7.029438018798828
165
+ ],
166
+ [
167
+ -21.609573364257812,
168
+ -40.83075714111328,
169
+ -21.334449768066406,
170
+ -46.67118835449219,
171
+ -7.869297027587891
172
+ ],
173
+ [
174
+ -23.818679809570312,
175
+ -44.688201904296875,
176
+ -24.13861083984375,
177
+ -50.31035614013672,
178
+ -8.673751831054688
179
+ ],
180
+ [
181
+ -26.139259338378906,
182
+ -47.874786376953125,
183
+ -26.761859893798828,
184
+ -54.02598571777344,
185
+ -9.537178039550781
186
+ ],
187
+ [
188
+ -28.473419189453125,
189
+ -50.6420783996582,
190
+ -29.38510513305664,
191
+ -57.71142578125,
192
+ -10.437923431396484
193
+ ],
194
+ [
195
+ -30.723190307617188,
196
+ -53.241661071777344,
197
+ -31.639423370361328,
198
+ -61.252540588378906,
199
+ -11.297477722167969
200
+ ],
201
+ [
202
+ -32.8468017578125,
203
+ -55.673526763916016,
204
+ -33.719932556152344,
205
+ -64.55354309082031,
206
+ -11.978679656982422
207
+ ],
208
+ [
209
+ -34.89494323730469,
210
+ -58.18925094604492,
211
+ -35.800437927246094,
212
+ -67.57942199707031,
213
+ -12.65576171875
214
+ ],
215
+ [
216
+ -36.79258728027344,
217
+ -60.70497131347656,
218
+ -37.790489196777344,
219
+ -70.3065185546875,
220
+ -13.115116119384766
221
+ ],
222
+ [
223
+ -38.41820526123047,
224
+ -63.38841247558594,
225
+ -39.96855163574219,
226
+ -72.76057434082031,
227
+ -13.479698181152344
228
+ ],
229
+ [
230
+ -39.82875061035156,
231
+ -65.73641967773438,
232
+ -42.222869873046875,
233
+ -74.99454498291016,
234
+ -14.06201171875
235
+ ],
236
+ [
237
+ -41.06019592285156,
238
+ -67.91671752929688,
239
+ -44.12246322631836,
240
+ -77.06887817382812,
241
+ -14.582843780517578
242
+ ],
243
+ [
244
+ -42.17378234863281,
245
+ -70.09700775146484,
246
+ -46.1196174621582,
247
+ -79.13003540039062,
248
+ -15.38583755493164
249
+ ]
250
+ ],
251
+ "max": [
252
+ [
253
+ 5.935976982116699,
254
+ 15.1451416015625,
255
+ 11.756964683532715,
256
+ 30.926345825195312,
257
+ 4.617164611816406
258
+ ],
259
+ [
260
+ 7.050804138183594,
261
+ 24.240886688232422,
262
+ 14.801684379577637,
263
+ 40.97989273071289,
264
+ 5.822669982910156
265
+ ],
266
+ [
267
+ 8.314274787902832,
268
+ 26.672752380371094,
269
+ 17.941551208496094,
270
+ 45.56070327758789,
271
+ 6.975105285644531
272
+ ],
273
+ [
274
+ 9.652067184448242,
275
+ 28.433757781982422,
276
+ 20.831390380859375,
277
+ 49.64745330810547,
278
+ 7.9514312744140625
279
+ ],
280
+ [
281
+ 10.692573547363281,
282
+ 30.027050018310547,
283
+ 25.21076202392578,
284
+ 53.44855499267578,
285
+ 8.888931274414062
286
+ ],
287
+ [
288
+ 11.881908416748047,
289
+ 31.36876678466797,
290
+ 29.00994873046875,
291
+ 57.183837890625,
292
+ 9.774349212646484
293
+ ],
294
+ [
295
+ 13.589406967163086,
296
+ 32.54277420043945,
297
+ 33.08050537109375,
298
+ 60.89848327636719,
299
+ 10.570022583007812
300
+ ],
301
+ [
302
+ 14.927953720092773,
303
+ 35.78352737426758,
304
+ 36.78923416137695,
305
+ 68.95306396484375,
306
+ 11.388931274414062
307
+ ],
308
+ [
309
+ 16.190017700195312,
310
+ 38.634681701660156,
311
+ 40.39845657348633,
312
+ 73.93795776367188,
313
+ 12.274349212646484
314
+ ],
315
+ [
316
+ 17.22235870361328,
317
+ 41.48583221435547,
318
+ 43.944515228271484,
319
+ 78.80449676513672,
320
+ 13.070022583007812
321
+ ],
322
+ [
323
+ 18.84430503845215,
324
+ 45.80079650878906,
325
+ 47.110504150390625,
326
+ 83.459228515625,
327
+ 13.78769302368164
328
+ ],
329
+ [
330
+ 20.551803588867188,
331
+ 49.082237243652344,
332
+ 50.16779327392578,
333
+ 87.70958709716797,
334
+ 14.449142456054688
335
+ ],
336
+ [
337
+ 22.117164611816406,
338
+ 52.18496322631836,
339
+ 53.71385192871094,
340
+ 91.4458236694336,
341
+ 15.10708236694336
342
+ ],
343
+ [
344
+ 23.676183700561523,
345
+ 55.36054611206055,
346
+ 56.553443908691406,
347
+ 94.54940032958984,
348
+ 15.787273406982422
349
+ ],
350
+ [
351
+ 25.160964965820312,
352
+ 58.64198684692383,
353
+ 58.729164123535156,
354
+ 97.79720306396484,
355
+ 16.412574768066406
356
+ ],
357
+ [
358
+ 26.52654266357422,
359
+ 61.636837005615234,
360
+ 60.90012741088867,
361
+ 102.08735656738281,
362
+ 17.037574768066406
363
+ ]
364
+ ],
365
+ "mean": [
366
+ [
367
+ -0.14882345497608185,
368
+ -0.49180787801742554,
369
+ -0.9774569869041443,
370
+ -0.7600352168083191,
371
+ -0.025079282000660896
372
+ ],
373
+ [
374
+ -0.2138441503047943,
375
+ -0.3800918459892273,
376
+ -0.9959100484848022,
377
+ -0.8829980492591858,
378
+ -0.030421346426010132
379
+ ],
380
+ [
381
+ -0.276848167181015,
382
+ -0.2752566337585449,
383
+ -1.0157051086425781,
384
+ -0.9944667816162109,
385
+ -0.03596708923578262
386
+ ],
387
+ [
388
+ -0.3380570411682129,
389
+ -0.17790287733078003,
390
+ -1.036616563796997,
391
+ -1.094215750694275,
392
+ -0.041609324514865875
393
+ ],
394
+ [
395
+ -0.3976344168186188,
396
+ -0.08703061938285828,
397
+ -1.0584867000579834,
398
+ -1.1839359998703003,
399
+ -0.04734227433800697
400
+ ],
401
+ [
402
+ -0.4557541310787201,
403
+ -0.0024509220384061337,
404
+ -1.0809900760650635,
405
+ -1.2642453908920288,
406
+ -0.053140584379434586
407
+ ],
408
+ [
409
+ -0.5126988887786865,
410
+ 0.07536793500185013,
411
+ -1.1036427021026611,
412
+ -1.3350884914398193,
413
+ -0.05906052142381668
414
+ ],
415
+ [
416
+ -0.5690915584564209,
417
+ 0.14643989503383636,
418
+ -1.1263598203659058,
419
+ -1.3966038227081299,
420
+ -0.06495664268732071
421
+ ],
422
+ [
423
+ -0.6256585121154785,
424
+ 0.21193456649780273,
425
+ -1.1488232612609863,
426
+ -1.4504973888397217,
427
+ -0.070767842233181
428
+ ],
429
+ [
430
+ -0.6831844449043274,
431
+ 0.2717616558074951,
432
+ -1.1708587408065796,
433
+ -1.4969099760055542,
434
+ -0.07657270133495331
435
+ ],
436
+ [
437
+ -0.741843044757843,
438
+ 0.3254620134830475,
439
+ -1.1922527551651,
440
+ -1.5359762907028198,
441
+ -0.08227676153182983
442
+ ],
443
+ [
444
+ -0.8015004992485046,
445
+ 0.37390172481536865,
446
+ -1.2126166820526123,
447
+ -1.568299412727356,
448
+ -0.08780235797166824
449
+ ],
450
+ [
451
+ -0.8629485368728638,
452
+ 0.41730016469955444,
453
+ -1.231200098991394,
454
+ -1.595727801322937,
455
+ -0.09321177750825882
456
+ ],
457
+ [
458
+ -0.9266560077667236,
459
+ 0.4554278254508972,
460
+ -1.2473944425582886,
461
+ -1.6191096305847168,
462
+ -0.09862378239631653
463
+ ],
464
+ [
465
+ -0.9926485419273376,
466
+ 0.4876198470592499,
467
+ -1.260882019996643,
468
+ -1.6385912895202637,
469
+ -0.10407595336437225
470
+ ],
471
+ [
472
+ -1.0612144470214844,
473
+ 0.5138142704963684,
474
+ -1.2710020542144775,
475
+ -1.6555445194244385,
476
+ -0.10952787846326828
477
+ ]
478
+ ],
479
+ "std": [
480
+ [
481
+ 0.577633798122406,
482
+ 2.619325876235962,
483
+ 1.7559900283813477,
484
+ 3.2910430431365967,
485
+ 0.709681510925293
486
+ ],
487
+ [
488
+ 1.0219744443893433,
489
+ 4.206506252288818,
490
+ 2.724515676498413,
491
+ 4.788137435913086,
492
+ 0.8845272064208984
493
+ ],
494
+ [
495
+ 1.5372892618179321,
496
+ 5.858143329620361,
497
+ 3.7363250255584717,
498
+ 6.292628765106201,
499
+ 1.057034969329834
500
+ ],
501
+ [
502
+ 2.0647077560424805,
503
+ 7.524584770202637,
504
+ 4.751458168029785,
505
+ 7.790938377380371,
506
+ 1.223911166191101
507
+ ],
508
+ [
509
+ 2.5927317142486572,
510
+ 9.188908576965332,
511
+ 5.757792949676514,
512
+ 9.275439262390137,
513
+ 1.3847688436508179
514
+ ],
515
+ [
516
+ 3.1178011894226074,
517
+ 10.844918251037598,
518
+ 6.75089168548584,
519
+ 10.744902610778809,
520
+ 1.5399470329284668
521
+ ],
522
+ [
523
+ 3.638444662094116,
524
+ 12.4906644821167,
525
+ 7.728830337524414,
526
+ 12.20102310180664,
527
+ 1.6899569034576416
528
+ ],
529
+ [
530
+ 4.152916431427002,
531
+ 14.12430191040039,
532
+ 8.691229820251465,
533
+ 13.647374153137207,
534
+ 1.835107445716858
535
+ ],
536
+ [
537
+ 4.659777641296387,
538
+ 15.742822647094727,
539
+ 9.637689590454102,
540
+ 15.077919006347656,
541
+ 1.9755820035934448
542
+ ],
543
+ [
544
+ 5.158328056335449,
545
+ 17.345365524291992,
546
+ 10.568361282348633,
547
+ 16.493101119995117,
548
+ 2.1115450859069824
549
+ ],
550
+ [
551
+ 5.648514270782471,
552
+ 18.93150520324707,
553
+ 11.483567237854004,
554
+ 17.894100189208984,
555
+ 2.2431843280792236
556
+ ],
557
+ [
558
+ 6.130801200866699,
559
+ 20.49928855895996,
560
+ 12.384008407592773,
561
+ 19.281808853149414,
562
+ 2.370608329772949
563
+ ],
564
+ [
565
+ 6.604080677032471,
566
+ 22.047456741333008,
567
+ 13.26936149597168,
568
+ 20.65220832824707,
569
+ 2.494211196899414
570
+ ],
571
+ [
572
+ 7.068251609802246,
573
+ 23.574750900268555,
574
+ 14.139976501464844,
575
+ 22.004636764526367,
576
+ 2.6142544746398926
577
+ ],
578
+ [
579
+ 7.523558139801025,
580
+ 25.080984115600586,
581
+ 14.996198654174805,
582
+ 23.340194702148438,
583
+ 2.7310373783111572
584
+ ],
585
+ [
586
+ 7.969391345977783,
587
+ 26.564414978027344,
588
+ 15.838000297546387,
589
+ 24.656681060791016,
590
+ 2.84470272064209
591
+ ]
592
+ ],
593
+ "q01": [
594
+ [
595
+ -1.8062204456329343,
596
+ -8.919847450256347,
597
+ -6.098232250213623,
598
+ -9.392142639160156,
599
+ -2.2635324096679685
600
+ ],
601
+ [
602
+ -2.9284115695953368,
603
+ -10.831075553894042,
604
+ -7.398877792358398,
605
+ -13.102965774536132,
606
+ -2.846776885986328
607
+ ],
608
+ [
609
+ -4.567752742767334,
610
+ -13.861269607543946,
611
+ -9.330749053955078,
612
+ -16.95085865020752,
613
+ -3.408384552001953
614
+ ],
615
+ [
616
+ -6.248722991943359,
617
+ -16.95799545288086,
618
+ -11.328928298950196,
619
+ -20.713608627319335,
620
+ -3.948393478393555
621
+ ],
622
+ [
623
+ -7.886484546661377,
624
+ -20.057677574157715,
625
+ -13.448420028686524,
626
+ -24.40064212799072,
627
+ -4.438659820556641
628
+ ],
629
+ [
630
+ -9.498039417266845,
631
+ -23.35751682281494,
632
+ -15.629821243286132,
633
+ -27.923939361572266,
634
+ -4.900402526855468
635
+ ],
636
+ [
637
+ -10.984527130126953,
638
+ -26.76651226043701,
639
+ -17.72079650878906,
640
+ -31.486430206298827,
641
+ -5.386660614013672
642
+ ],
643
+ [
644
+ -12.448093795776368,
645
+ -30.02010570526123,
646
+ -19.850285949707033,
647
+ -34.944140396118165,
648
+ -5.8193305969238285
649
+ ],
650
+ [
651
+ -13.897714653015136,
652
+ -33.15575813293457,
653
+ -21.75669013977051,
654
+ -38.49285972595215,
655
+ -6.281961669921875
656
+ ],
657
+ [
658
+ -15.275969734191895,
659
+ -36.156199798583984,
660
+ -23.685923233032227,
661
+ -41.66505508422851,
662
+ -6.699819717407227
663
+ ],
664
+ [
665
+ -16.649525146484375,
666
+ -39.16726852416992,
667
+ -25.531164245605467,
668
+ -44.942562408447266,
669
+ -7.123175735473633
670
+ ],
671
+ [
672
+ -18.179378204345703,
673
+ -42.189204406738284,
674
+ -27.486108474731445,
675
+ -48.14005401611328,
676
+ -7.559758148193359
677
+ ],
678
+ [
679
+ -19.6202645111084,
680
+ -45.06910652160644,
681
+ -29.289424438476562,
682
+ -51.29126266479492,
683
+ -7.924339752197265
684
+ ],
685
+ [
686
+ -21.03468589782715,
687
+ -48.01726432800293,
688
+ -31.032142181396484,
689
+ -54.37339141845703,
690
+ -8.310578994750976
691
+ ],
692
+ [
693
+ -22.37120086669922,
694
+ -50.86057395935059,
695
+ -32.76242889404297,
696
+ -57.4203475189209,
697
+ -8.610725784301758
698
+ ],
699
+ [
700
+ -23.674882736206055,
701
+ -53.774470825195316,
702
+ -34.39431442260742,
703
+ -60.04805870056152,
704
+ -8.919435577392578
705
+ ]
706
+ ],
707
+ "q99": [
708
+ [
709
+ 1.4155050659179678,
710
+ 5.832528152465819,
711
+ 3.770242309570312,
712
+ 8.264584426879878,
713
+ 2.388116912841793
714
+ ],
715
+ [
716
+ 2.2402151679992675,
717
+ 8.16798316955566,
718
+ 6.776809844970698,
719
+ 12.389562988281249,
720
+ 2.9987688446044904
721
+ ],
722
+ [
723
+ 3.5463735818862903,
724
+ 11.233944664001463,
725
+ 9.727340927124018,
726
+ 16.413600959777828,
727
+ 3.604117355346678
728
+ ],
729
+ [
730
+ 4.668624420166013,
731
+ 14.407206115722651,
732
+ 12.673404464721656,
733
+ 20.60697456359862,
734
+ 4.160473480224606
735
+ ],
736
+ [
737
+ 5.865899009704584,
738
+ 17.624989166259766,
739
+ 15.56380615234375,
740
+ 24.551775817871093,
741
+ 4.659558868408203
742
+ ],
743
+ [
744
+ 7.058602085113526,
745
+ 20.7240623474121,
746
+ 18.425717315673825,
747
+ 28.903741188049313,
748
+ 5.166472625732418
749
+ ],
750
+ [
751
+ 8.227654857635498,
752
+ 23.790639610290516,
753
+ 21.367591400146477,
754
+ 32.793743743896485,
755
+ 5.631255874633785
756
+ ],
757
+ [
758
+ 9.443681087493887,
759
+ 26.656646537780755,
760
+ 24.202687225341794,
761
+ 36.56039634704586,
762
+ 6.100628433227524
763
+ ],
764
+ [
765
+ 10.52673578262327,
766
+ 29.699559097290038,
767
+ 26.936041069030754,
768
+ 40.60598495483394,
769
+ 6.594990081787106
770
+ ],
771
+ [
772
+ 11.641950950622554,
773
+ 32.64166328430176,
774
+ 29.66598850250244,
775
+ 44.52786010742182,
776
+ 6.949279327392576
777
+ ],
778
+ [
779
+ 12.696785278320302,
780
+ 35.440256652832026,
781
+ 32.34731643676756,
782
+ 48.332406845092756,
783
+ 7.380425491332997
784
+ ],
785
+ [
786
+ 13.630289459228491,
787
+ 38.30537055969238,
788
+ 34.77291099548339,
789
+ 52.09018577575682,
790
+ 7.733721542358392
791
+ ],
792
+ [
793
+ 14.566434345245357,
794
+ 41.147760696411126,
795
+ 37.24289001464843,
796
+ 55.61033187866211,
797
+ 8.013584899902341
798
+ ],
799
+ [
800
+ 15.468500289916962,
801
+ 43.97983375549314,
802
+ 39.60062606811521,
803
+ 58.937206420898434,
804
+ 8.37797401428222
805
+ ],
806
+ [
807
+ 16.33366302490234,
808
+ 46.76957221984862,
809
+ 42.0470317840576,
810
+ 62.48077819824218,
811
+ 8.64075592041015
812
+ ],
813
+ [
814
+ 17.328776473999017,
815
+ 49.3529705047607,
816
+ 44.19434700012206,
817
+ 65.80092971801756,
818
+ 8.837472381591791
819
+ ]
820
+ ]
821
+ }
822
+ }
823
+ }
824
+ }
experiment_cfg/final_model_config.json ADDED
@@ -0,0 +1,53 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ "apply_sincos_state_encoding": true,
19
+ "use_relative_action": true,
20
+ "max_state_dim": 128,
21
+ "max_action_dim": 128,
22
+ "action_horizon": 50,
23
+ "hidden_size": 1024,
24
+ "input_embedding_dim": 1536,
25
+ "add_pos_embed": true,
26
+ "attn_dropout": 0.2,
27
+ "use_vlln": true,
28
+ "max_seq_len": 1024,
29
+ "use_alternate_vl_dit": true,
30
+ "attend_text_every_n_blocks": 2,
31
+ "diffusion_model_cfg": {
32
+ "attention_head_dim": 48,
33
+ "dropout": 0.2,
34
+ "final_dropout": true,
35
+ "interleave_self_attention": true,
36
+ "norm_type": "ada_norm",
37
+ "num_attention_heads": 32,
38
+ "num_layers": 32,
39
+ "output_dim": 1024,
40
+ "positional_embeddings": null
41
+ },
42
+ "num_inference_timesteps": 4,
43
+ "noise_beta_alpha": 1.5,
44
+ "noise_beta_beta": 1.0,
45
+ "noise_s": 0.999,
46
+ "num_timestep_buckets": 1000,
47
+ "tune_projector": true,
48
+ "tune_diffusion_model": true,
49
+ "tune_vlln": true,
50
+ "state_dropout_prob": 0.0,
51
+ "state_additive_noise_scale": 0.0,
52
+ "max_num_embodiments": 32
53
+ }
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 sha256:75ddfd63629c7ed28d9e115cf597eb9d2cc244882f3ace7837792074714aa033
3
+ size 4990120184
model-00002-of-00002.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d37cb96c3898e40ba5fcba0b13fda3faa441ba68fb6de007f34d0fedec73e99
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:ee445bdaa0397e6915437c63ea599569eea52538b0ac93d52e00766e2f5dba75
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"
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:f40f14c749486cb921a4172109eec747f70a02e7084acf7a22edd8d56f985a31
3
+ size 14645
scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d02be6d8bda4ea9c67040ed89f878acdc986bd4df3fbc60440a9d3eacca02d63
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 sha256:2c9d136fa72f1f85b917a8c017de030ccb450d50b83aab88ed725d6d9dec57c7
3
+ size 5841
wandb_config.json ADDED
@@ -0,0 +1 @@
 
 
1
+ {"project": "finetune-gr00t-n1d6", "run_id": "grootn16-finetune_sreetz-so101_teleop_vials_rack_left_sim_and_real"}