Robotics
LeRobot
Safetensors
gr00t_n1d6
nvkartik commited on
Commit
e622edd
·
verified ·
1 Parent(s): 3a3179e

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +137 -0
  3. model.safetensors +3 -0
  4. train_config.json +272 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: izuluaga/finish_sandwich
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: gr00t_n1d6
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - lerobot
9
+ - gr00t_n1d6
10
+ - robotics
11
+ ---
12
+
13
+ # Model Card for gr00t_n1d6
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+ _Model type not recognized — please update this template._
19
+
20
+
21
+ This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
22
+ See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).
23
+
24
+ ---
25
+
26
+ ## How to Get Started with the Model
27
+
28
+ For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
29
+ Below is the short version on how to train and run inference/eval:
30
+
31
+ ### Train from scratch
32
+
33
+ ```bash
34
+ lerobot-train \
35
+ --dataset.repo_id=${HF_USER}/<dataset> \
36
+ --policy.type=act \
37
+ --output_dir=outputs/train/<desired_policy_repo_id> \
38
+ --job_name=lerobot_training \
39
+ --policy.device=cuda \
40
+ --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
41
+ --wandb.enable=true
42
+ ```
43
+
44
+ _Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`._
45
+
46
+ ### Evaluate the policy/run inference
47
+
48
+ ```bash
49
+ lerobot-record \
50
+ --robot.type=so100_follower \
51
+ --dataset.repo_id=<hf_user>/eval_<dataset> \
52
+ --policy.path=<hf_user>/<desired_policy_repo_id> \
53
+ --episodes=10
54
+ ```
55
+
56
+ Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.
57
+
58
+ ---
59
+
60
+ ## Model Details
61
+
62
+ - **License:** apache-2.0
config.json ADDED
@@ -0,0 +1,137 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "gr00t_n1d6",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 6
9
+ ]
10
+ },
11
+ "observation.images.wrist": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 480,
16
+ 640
17
+ ]
18
+ },
19
+ "observation.images.front": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 480,
24
+ 640
25
+ ]
26
+ }
27
+ },
28
+ "output_features": {
29
+ "action": {
30
+ "type": "ACTION",
31
+ "shape": [
32
+ 6
33
+ ]
34
+ }
35
+ },
36
+ "device": "cuda",
37
+ "use_amp": false,
38
+ "push_to_hub": true,
39
+ "repo_id": "nvkartik/gr00t_n1d6-finish_sandwich-abs-action-true-tune",
40
+ "private": null,
41
+ "tags": null,
42
+ "license": null,
43
+ "pretrained_path": null,
44
+ "chunk_size": 40,
45
+ "n_action_steps": 40,
46
+ "max_state_dim": 29,
47
+ "max_action_dim": 29,
48
+ "normalization_mapping": {
49
+ "VISUAL": "IDENTITY",
50
+ "STATE": "MEAN_STD",
51
+ "ACTION": "MEAN_STD"
52
+ },
53
+ "image_size": [
54
+ 224,
55
+ 224
56
+ ],
57
+ "base_model_path": "nvidia/GR00T-N1.6-3B",
58
+ "tokenizer_assets_repo": "nvidia/Eagle-Block2A-2B-v2",
59
+ "backbone_model_type": "eagle",
60
+ "model_revision": null,
61
+ "backbone_embedding_dim": 2048,
62
+ "select_layer": 16,
63
+ "reproject_vision": false,
64
+ "use_flash_attention": true,
65
+ "load_bf16": true,
66
+ "eagle_collator": false,
67
+ "backbone_trainable_params_fp32": true,
68
+ "embodiment_tag": "new_embodiment",
69
+ "tune_top_llm_layers": 4,
70
+ "tune_llm": false,
71
+ "tune_visual": false,
72
+ "tune_projector": true,
73
+ "tune_diffusion_model": true,
74
+ "tune_vlln": true,
75
+ "image_crop_size": null,
76
+ "image_target_size": null,
77
+ "shortest_image_edge": 256,
78
+ "crop_fraction": 0.95,
79
+ "random_rotation_angle": null,
80
+ "color_jitter_params": null,
81
+ "use_albumentations_transforms": true,
82
+ "formalize_language": true,
83
+ "lora_rank": 0,
84
+ "lora_alpha": 16,
85
+ "lora_dropout": 0.1,
86
+ "lora_full_model": false,
87
+ "action_horizon": 16,
88
+ "hidden_size": 1024,
89
+ "input_embedding_dim": 1536,
90
+ "add_pos_embed": true,
91
+ "attn_dropout": 0.2,
92
+ "use_vlln": true,
93
+ "max_seq_len": 1024,
94
+ "use_alternate_vl_dit": true,
95
+ "attend_text_every_n_blocks": 2,
96
+ "diffusion_model_cfg": {
97
+ "positional_embeddings": null,
98
+ "num_layers": 32,
99
+ "num_attention_heads": 32,
100
+ "attention_head_dim": 48,
101
+ "norm_type": "ada_norm",
102
+ "dropout": 0.2,
103
+ "final_dropout": true,
104
+ "output_dim": 1024,
105
+ "interleave_self_attention": true
106
+ },
107
+ "num_inference_timesteps": 4,
108
+ "noise_beta_alpha": 1.5,
109
+ "noise_beta_beta": 1.0,
110
+ "noise_s": 0.999,
111
+ "num_timestep_buckets": 1000,
112
+ "state_dropout_prob": 0.0,
113
+ "state_additive_noise_scale": 0.0,
114
+ "apply_sincos_state_encoding": false,
115
+ "use_relative_action": false,
116
+ "max_num_embodiments": 32,
117
+ "optimizer_lr": 0.0001,
118
+ "optimizer_betas": [
119
+ 0.95,
120
+ 0.999
121
+ ],
122
+ "optimizer_eps": 1e-08,
123
+ "optimizer_weight_decay": 1e-05,
124
+ "warmup_ratio": 0.05,
125
+ "use_bf16": true,
126
+ "video_backend": "decord",
127
+ "balance_dataset_weights": true,
128
+ "balance_trajectory_weights": true,
129
+ "dataset_paths": null,
130
+ "output_dir": "./tmp/gr00t_n16",
131
+ "save_steps": 1000,
132
+ "max_steps": 10000,
133
+ "batch_size": 32,
134
+ "dataloader_num_workers": 8,
135
+ "report_to": "wandb",
136
+ "resume": false
137
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b2197bf0e2bf6c7211d5add96c0c33c8bc21f7d7708e4882c0742ebd7a0f6101
3
+ size 9192043768
train_config.json ADDED
@@ -0,0 +1,272 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "izuluaga/finish_sandwich",
4
+ "root": null,
5
+ "episodes": null,
6
+ "image_transforms": {
7
+ "enable": false,
8
+ "max_num_transforms": 3,
9
+ "random_order": false,
10
+ "tfs": {
11
+ "brightness": {
12
+ "weight": 1.0,
13
+ "type": "ColorJitter",
14
+ "kwargs": {
15
+ "brightness": [
16
+ 0.8,
17
+ 1.2
18
+ ]
19
+ }
20
+ },
21
+ "contrast": {
22
+ "weight": 1.0,
23
+ "type": "ColorJitter",
24
+ "kwargs": {
25
+ "contrast": [
26
+ 0.8,
27
+ 1.2
28
+ ]
29
+ }
30
+ },
31
+ "saturation": {
32
+ "weight": 1.0,
33
+ "type": "ColorJitter",
34
+ "kwargs": {
35
+ "saturation": [
36
+ 0.5,
37
+ 1.5
38
+ ]
39
+ }
40
+ },
41
+ "hue": {
42
+ "weight": 1.0,
43
+ "type": "ColorJitter",
44
+ "kwargs": {
45
+ "hue": [
46
+ -0.05,
47
+ 0.05
48
+ ]
49
+ }
50
+ },
51
+ "sharpness": {
52
+ "weight": 1.0,
53
+ "type": "SharpnessJitter",
54
+ "kwargs": {
55
+ "sharpness": [
56
+ 0.5,
57
+ 1.5
58
+ ]
59
+ }
60
+ },
61
+ "affine": {
62
+ "weight": 1.0,
63
+ "type": "RandomAffine",
64
+ "kwargs": {
65
+ "degrees": [
66
+ -5.0,
67
+ 5.0
68
+ ],
69
+ "translate": [
70
+ 0.05,
71
+ 0.05
72
+ ]
73
+ }
74
+ }
75
+ }
76
+ },
77
+ "revision": null,
78
+ "use_imagenet_stats": true,
79
+ "video_backend": "torchcodec",
80
+ "streaming": false
81
+ },
82
+ "env": null,
83
+ "policy": {
84
+ "type": "gr00t_n1d6",
85
+ "n_obs_steps": 1,
86
+ "input_features": {
87
+ "observation.state": {
88
+ "type": "STATE",
89
+ "shape": [
90
+ 6
91
+ ]
92
+ },
93
+ "observation.images.wrist": {
94
+ "type": "VISUAL",
95
+ "shape": [
96
+ 3,
97
+ 480,
98
+ 640
99
+ ]
100
+ },
101
+ "observation.images.front": {
102
+ "type": "VISUAL",
103
+ "shape": [
104
+ 3,
105
+ 480,
106
+ 640
107
+ ]
108
+ }
109
+ },
110
+ "output_features": {
111
+ "action": {
112
+ "type": "ACTION",
113
+ "shape": [
114
+ 6
115
+ ]
116
+ }
117
+ },
118
+ "device": "cuda",
119
+ "use_amp": false,
120
+ "push_to_hub": true,
121
+ "repo_id": "nvkartik/gr00t_n1d6-finish_sandwich-abs-action-true-tune",
122
+ "private": null,
123
+ "tags": null,
124
+ "license": null,
125
+ "pretrained_path": null,
126
+ "chunk_size": 40,
127
+ "n_action_steps": 40,
128
+ "max_state_dim": 29,
129
+ "max_action_dim": 29,
130
+ "normalization_mapping": {
131
+ "VISUAL": "IDENTITY",
132
+ "STATE": "MEAN_STD",
133
+ "ACTION": "MEAN_STD"
134
+ },
135
+ "image_size": [
136
+ 224,
137
+ 224
138
+ ],
139
+ "base_model_path": "nvidia/GR00T-N1.6-3B",
140
+ "tokenizer_assets_repo": "nvidia/Eagle-Block2A-2B-v2",
141
+ "backbone_model_type": "eagle",
142
+ "model_revision": null,
143
+ "backbone_embedding_dim": 2048,
144
+ "select_layer": 16,
145
+ "reproject_vision": false,
146
+ "use_flash_attention": true,
147
+ "load_bf16": true,
148
+ "eagle_collator": false,
149
+ "backbone_trainable_params_fp32": true,
150
+ "embodiment_tag": "new_embodiment",
151
+ "tune_top_llm_layers": 4,
152
+ "tune_llm": false,
153
+ "tune_visual": false,
154
+ "tune_projector": true,
155
+ "tune_diffusion_model": true,
156
+ "tune_vlln": true,
157
+ "image_crop_size": null,
158
+ "image_target_size": null,
159
+ "shortest_image_edge": 256,
160
+ "crop_fraction": 0.95,
161
+ "random_rotation_angle": null,
162
+ "color_jitter_params": null,
163
+ "use_albumentations_transforms": true,
164
+ "formalize_language": true,
165
+ "lora_rank": 0,
166
+ "lora_alpha": 16,
167
+ "lora_dropout": 0.1,
168
+ "lora_full_model": false,
169
+ "action_horizon": 16,
170
+ "hidden_size": 1024,
171
+ "input_embedding_dim": 1536,
172
+ "add_pos_embed": true,
173
+ "attn_dropout": 0.2,
174
+ "use_vlln": true,
175
+ "max_seq_len": 1024,
176
+ "use_alternate_vl_dit": true,
177
+ "attend_text_every_n_blocks": 2,
178
+ "diffusion_model_cfg": {
179
+ "positional_embeddings": null,
180
+ "num_layers": 32,
181
+ "num_attention_heads": 32,
182
+ "attention_head_dim": 48,
183
+ "norm_type": "ada_norm",
184
+ "dropout": 0.2,
185
+ "final_dropout": true,
186
+ "output_dim": 1024,
187
+ "interleave_self_attention": true
188
+ },
189
+ "num_inference_timesteps": 4,
190
+ "noise_beta_alpha": 1.5,
191
+ "noise_beta_beta": 1.0,
192
+ "noise_s": 0.999,
193
+ "num_timestep_buckets": 1000,
194
+ "state_dropout_prob": 0.0,
195
+ "state_additive_noise_scale": 0.0,
196
+ "apply_sincos_state_encoding": false,
197
+ "use_relative_action": false,
198
+ "max_num_embodiments": 32,
199
+ "optimizer_lr": 0.0001,
200
+ "optimizer_betas": [
201
+ 0.95,
202
+ 0.999
203
+ ],
204
+ "optimizer_eps": 1e-08,
205
+ "optimizer_weight_decay": 1e-05,
206
+ "warmup_ratio": 0.05,
207
+ "use_bf16": true,
208
+ "video_backend": "decord",
209
+ "balance_dataset_weights": true,
210
+ "balance_trajectory_weights": true,
211
+ "dataset_paths": null,
212
+ "output_dir": "./tmp/gr00t_n16",
213
+ "save_steps": 1000,
214
+ "max_steps": 10000,
215
+ "batch_size": 32,
216
+ "dataloader_num_workers": 8,
217
+ "report_to": "wandb",
218
+ "resume": false
219
+ },
220
+ "output_dir": "outputs/train/gr00t_n1d6-finish_sandwich-20251230-002210",
221
+ "job_name": "gr00t_n1d6-finish_sandwich-20251230-002210",
222
+ "resume": false,
223
+ "seed": 1000,
224
+ "num_workers": 4,
225
+ "batch_size": 32,
226
+ "steps": 20000,
227
+ "eval_freq": 20000,
228
+ "log_freq": 50,
229
+ "tolerance_s": 0.0001,
230
+ "save_checkpoint": true,
231
+ "save_freq": 5000,
232
+ "use_policy_training_preset": true,
233
+ "optimizer": {
234
+ "type": "adamw",
235
+ "lr": 0.0001,
236
+ "weight_decay": 1e-05,
237
+ "grad_clip_norm": 10.0,
238
+ "betas": [
239
+ 0.95,
240
+ 0.999
241
+ ],
242
+ "eps": 1e-08
243
+ },
244
+ "scheduler": {
245
+ "type": "cosine_decay_with_warmup",
246
+ "num_warmup_steps": 500,
247
+ "num_decay_steps": 10000,
248
+ "peak_lr": 0.0001,
249
+ "decay_lr": 1e-05
250
+ },
251
+ "eval": {
252
+ "n_episodes": 50,
253
+ "batch_size": 50,
254
+ "use_async_envs": false
255
+ },
256
+ "wandb": {
257
+ "enable": true,
258
+ "disable_artifact": false,
259
+ "project": "lerobot",
260
+ "entity": null,
261
+ "notes": null,
262
+ "run_id": "4ktud30e",
263
+ "mode": null
264
+ },
265
+ "use_rabc": false,
266
+ "rabc_progress_path": null,
267
+ "rabc_kappa": 0.01,
268
+ "rabc_epsilon": 1e-06,
269
+ "rabc_head_mode": "sparse",
270
+ "rename_map": {},
271
+ "checkpoint_path": null
272
+ }