Robotics
LeRobot
Safetensors
groot
sreetz-nv commited on
Commit
281cf42
·
verified ·
1 Parent(s): 7f8f255

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +167 -0
  2. config.json +100 -0
  3. model.safetensors +3 -0
  4. train_config.json +245 -0
README.md ADDED
@@ -0,0 +1,167 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: sreetz-nv/so101-clean-up-vials-into-rack-50_20260628_131121
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: groot
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - groot
9
+ - lerobot
10
+ - robotics
11
+ ---
12
+
13
+ # Model Card for groot
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+ [GR00T N1.5](https://github.com/NVIDIA/Isaac-GR00T) is an open, cross-embodiment foundation model from NVIDIA for generalized humanoid robot reasoning and skills. It takes language and images as input and uses a flow-matching action transformer to predict actions conditioned on vision, language, and proprioception.
19
+
20
+
21
+
22
+ <p align="center">
23
+ <img src="https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/lerobot/lerobot-groot-paper1%20(1).png" alt="groot architecture" width="85%"/>
24
+ </p>
25
+
26
+
27
+ <!-- A short demo is worth more than any description! Record a GIF/video of the policy
28
+ running on your robot, upload it to this repo, and embed it here:
29
+ <p align="center">
30
+ <img src="https://huggingface.co/<hf_user>/<policy_repo_id>/resolve/main/demo.gif" width="60%"/>
31
+ </p>
32
+ -->
33
+
34
+ This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
35
+
36
+ Learn how to train and run it in the [LeRobot groot guide](https://huggingface.co/docs/lerobot/main/en/groot), or browse the [full documentation](https://huggingface.co/docs/lerobot/index).
37
+
38
+
39
+ ---
40
+
41
+ ## Model Details
42
+
43
+ - **License:** apache-2.0
44
+ - **Robot type:** `so_follower`
45
+ - **Cameras:** `wrist`, `front`
46
+
47
+
48
+ ## Inputs & Outputs
49
+
50
+ The policy consumes these observation features and produces these action features.
51
+
52
+ **Inputs**
53
+
54
+ | Feature | Type | Shape |
55
+ | --- | --- | --- |
56
+ | `observation.state` | STATE | `(6,)` |
57
+ | `observation.images.wrist` | VISUAL | `(3, 480, 640)` |
58
+ | `observation.images.front` | VISUAL | `(3, 480, 640)` |
59
+
60
+ **Outputs**
61
+
62
+ | Feature | Type | Shape |
63
+ | --- | --- | --- |
64
+ | `action` | ACTION | `(6,)` |
65
+
66
+
67
+ ## Training Dataset
68
+
69
+ - **Repository:** [sreetz-nv/so101-clean-up-vials-into-rack-50_20260628_131121](https://huggingface.co/datasets/sreetz-nv/so101-clean-up-vials-into-rack-50_20260628_131121)
70
+ - **Episodes:** 50
71
+ - **Frames:** 52304
72
+ - **Frame rate:** 30 FPS
73
+ - **Task(s):** "Pick up the vials and put them in the yellow rack"
74
+
75
+ <a class="flex" href="https://huggingface.co/spaces/lerobot/visualize_dataset?path=sreetz-nv/so101-clean-up-vials-into-rack-50_20260628_131121">
76
+ <img class="block dark:hidden" src="https://huggingface.co/datasets/huggingface/badges/resolve/main/visualize-this-dataset-xl.svg"/>
77
+ <img class="hidden dark:block" src="https://huggingface.co/datasets/huggingface/badges/resolve/main/visualize-this-dataset-xl-dark.svg"/>
78
+ </a>
79
+
80
+
81
+ ## Training Configuration
82
+
83
+ | Setting | Value |
84
+ | --- | --- |
85
+ | Training steps | 20000 |
86
+ | Batch size | 64 |
87
+ | Optimizer | adamw |
88
+ | Learning rate | 0.0001 |
89
+ | Seed | 42 |
90
+ | LeRobot version | 0.5.2 |
91
+
92
+ ---
93
+
94
+ ## How to Get Started with the Model
95
+
96
+ New to LeRobot? These guides cover the full workflow:
97
+
98
+ - **[Install LeRobot](https://huggingface.co/docs/lerobot/main/en/installation)** — set up the `lerobot` package.
99
+ - **[Hardware setup](https://huggingface.co/docs/lerobot/main/en/hardware_guide)** — assemble, wire, and calibrate your robot and cameras.
100
+ - **[Record data & train a policy](https://huggingface.co/docs/lerobot/en/il_robots)** — the end-to-end imitation-learning walkthrough.
101
+ - **[CLI cheat-sheet](https://huggingface.co/docs/lerobot/main/en/cheat-sheet)** — quick reference for the `lerobot-*` commands.
102
+
103
+ The short version to run and train this policy:
104
+
105
+ ### Run the policy on your robot
106
+
107
+ ```bash
108
+ lerobot-rollout \
109
+ --strategy.type=base \
110
+ --robot.type=so_follower \
111
+ --robot.port=<your_robot_port> \
112
+ --robot.cameras="{ <camera_1>: {type: opencv, index_or_path: <index_or_path>, width: 640, height: 480, fps: 30}, <camera_2>: {type: opencv, index_or_path: <index_or_path>, width: 640, height: 480, fps: 30}}" \
113
+ --policy.path=sreetz-nv/so101-clean-up-vials-into-rack-50-groot-n17-relact-bs64-20k-20260628_lerobot_fix \
114
+ --task="Pick up the vials and put them in the yellow rack" \
115
+ --duration=60
116
+ ```
117
+
118
+ Replace the remaining `<...>` placeholders with your own values: `--robot.port` and the camera names/indices are specific to your machine, and the camera names must match the observation keys this policy was trained on.
119
+
120
+ When `--strategy.type=base` is used the script doesn't record the episodes. Skipping duration will make the policy run indefinitely. For more information look at [rollout documentation](https://huggingface.co/docs/lerobot/main/en/inference).
121
+
122
+ ### Train your own policy
123
+
124
+ ```bash
125
+ lerobot-train \
126
+ --dataset.repo_id=${HF_USER}/<dataset> \
127
+ --policy.type=groot \
128
+ --output_dir=outputs/train/<policy_repo_id> \
129
+ --job_name=lerobot_training \
130
+ --policy.device=cuda \
131
+ --policy.repo_id=${HF_USER}/<policy_repo_id> \
132
+ --wandb.enable=true
133
+ ```
134
+
135
+ _Writes checkpoints to `outputs/train/<policy_repo_id>/checkpoints/`._
136
+
137
+ ---
138
+
139
+ ## Evaluation
140
+
141
+ <!-- Report real-robot results here: run the policy several times per task and count the
142
+ successes. Delete the "No evaluation results" line and fill in this table instead:
143
+
144
+ | Task | Trials | Successes | Success rate |
145
+ | ---- | ------ | --------- | ------------ |
146
+ | pick the lego brick | 10 | 8 | 80% |
147
+
148
+ Also worth noting: anything that affects difficulty (new object positions, lighting,
149
+ distractors, a different robot of the same type, ...).
150
+ -->
151
+
152
+ _No evaluation results have been provided for this policy yet._
153
+
154
+ ---
155
+
156
+ ## Citation
157
+
158
+ If you use this policy, please cite the method linked in the description above, along with LeRobot:
159
+
160
+ ```bibtex
161
+ @misc{cadene2024lerobot,
162
+ author = {Cadene, Remi and Alibert, Simon and Soare, Alexander and Gallouedec, Quentin and Zouitine, Adil and Palma, Steven and Kooijmans, Pepijn and Aractingi, Michel and Shukor, Mustafa and Aubakirova, Dana and Russi, Martino and Capuano, Francesco and Pascal, Caroline and Choghari, Jade and Moss, Jess and Wolf, Thomas},
163
+ title = {LeRobot: State-of-the-art Machine Learning for Real-World Robotics in Pytorch},
164
+ howpublished = "\url{https://github.com/huggingface/lerobot}",
165
+ year = {2024}
166
+ }
167
+ ```
config.json ADDED
@@ -0,0 +1,100 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "groot",
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
+ "use_peft": false,
39
+ "push_to_hub": true,
40
+ "repo_id": "sreetz-nv/so101-clean-up-vials-into-rack-50-groot-n17-relact-bs64-20k-20260628_lerobot_fix",
41
+ "private": null,
42
+ "tags": null,
43
+ "license": null,
44
+ "pretrained_path": "outputs/train/sreetz-nv/so101-clean-up-vials-into-rack-50-groot-n17-relact-bs64-20k-20260628_lerobot_fix/checkpoints/last/pretrained_model",
45
+ "pretrained_revision": null,
46
+ "chunk_size": 16,
47
+ "n_action_steps": 16,
48
+ "max_state_dim": 132,
49
+ "max_action_dim": 132,
50
+ "normalization_mapping": {
51
+ "VISUAL": "IDENTITY",
52
+ "STATE": "IDENTITY",
53
+ "ACTION": "IDENTITY"
54
+ },
55
+ "base_model_path": "nvidia/GR00T-N1.7-3B",
56
+ "action_decode_transform": null,
57
+ "embodiment_tag": "new_embodiment",
58
+ "tune_llm": false,
59
+ "tune_visual": false,
60
+ "tune_projector": true,
61
+ "tune_diffusion_model": true,
62
+ "tune_vlln": true,
63
+ "tune_top_llm_layers": 0,
64
+ "num_inference_timesteps": null,
65
+ "rtc_ramp_rate": null,
66
+ "use_flash_attention": true,
67
+ "use_relative_actions": true,
68
+ "relative_exclude_joints": [
69
+ "gripper"
70
+ ],
71
+ "optimizer_lr": 0.0001,
72
+ "optimizer_betas": [
73
+ 0.95,
74
+ 0.999
75
+ ],
76
+ "optimizer_eps": 1e-08,
77
+ "optimizer_weight_decay": 1e-05,
78
+ "warmup_ratio": 0.05,
79
+ "use_bf16": true,
80
+ "image_size": [
81
+ 256,
82
+ 256
83
+ ],
84
+ "tokenizer_assets_repo": null,
85
+ "lora_rank": 0,
86
+ "lora_alpha": 16,
87
+ "lora_dropout": 0.1,
88
+ "lora_full_model": false,
89
+ "video_backend": "decord",
90
+ "balance_dataset_weights": true,
91
+ "balance_trajectory_weights": true,
92
+ "dataset_paths": null,
93
+ "output_dir": "./tmp/gr00t",
94
+ "save_steps": 1000,
95
+ "max_steps": 10000,
96
+ "batch_size": 32,
97
+ "dataloader_num_workers": 8,
98
+ "report_to": "wandb",
99
+ "resume": false
100
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:91c501404abc064e9e839da4357a4bcde0145379ba703a8205137c35f60c0ac8
3
+ size 10579843072
train_config.json ADDED
@@ -0,0 +1,245 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "sreetz-nv/so101-clean-up-vials-into-rack-50_20260628_131121",
4
+ "root": null,
5
+ "episodes": null,
6
+ "image_transforms": {
7
+ "enable": true,
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
+ "return_uint8": false,
81
+ "depth_output_unit": "mm",
82
+ "streaming": false,
83
+ "eval_split": 0.0
84
+ },
85
+ "env": null,
86
+ "policy": {
87
+ "type": "groot",
88
+ "n_obs_steps": 1,
89
+ "input_features": {
90
+ "observation.state": {
91
+ "type": "STATE",
92
+ "shape": [
93
+ 6
94
+ ]
95
+ },
96
+ "observation.images.wrist": {
97
+ "type": "VISUAL",
98
+ "shape": [
99
+ 3,
100
+ 480,
101
+ 640
102
+ ]
103
+ },
104
+ "observation.images.front": {
105
+ "type": "VISUAL",
106
+ "shape": [
107
+ 3,
108
+ 480,
109
+ 640
110
+ ]
111
+ }
112
+ },
113
+ "output_features": {
114
+ "action": {
115
+ "type": "ACTION",
116
+ "shape": [
117
+ 6
118
+ ]
119
+ }
120
+ },
121
+ "device": "cuda",
122
+ "use_amp": false,
123
+ "use_peft": false,
124
+ "push_to_hub": true,
125
+ "repo_id": "sreetz-nv/so101-clean-up-vials-into-rack-50-groot-n17-relact-bs64-20k-20260628_lerobot_fix",
126
+ "private": null,
127
+ "tags": null,
128
+ "license": null,
129
+ "pretrained_path": "outputs/train/sreetz-nv/so101-clean-up-vials-into-rack-50-groot-n17-relact-bs64-20k-20260628_lerobot_fix/checkpoints/last/pretrained_model",
130
+ "pretrained_revision": null,
131
+ "chunk_size": 16,
132
+ "n_action_steps": 16,
133
+ "max_state_dim": 132,
134
+ "max_action_dim": 132,
135
+ "normalization_mapping": {
136
+ "VISUAL": "IDENTITY",
137
+ "STATE": "IDENTITY",
138
+ "ACTION": "IDENTITY"
139
+ },
140
+ "base_model_path": "nvidia/GR00T-N1.7-3B",
141
+ "action_decode_transform": null,
142
+ "embodiment_tag": "new_embodiment",
143
+ "tune_llm": false,
144
+ "tune_visual": false,
145
+ "tune_projector": true,
146
+ "tune_diffusion_model": true,
147
+ "tune_vlln": true,
148
+ "tune_top_llm_layers": 0,
149
+ "num_inference_timesteps": null,
150
+ "rtc_ramp_rate": null,
151
+ "use_flash_attention": true,
152
+ "use_relative_actions": true,
153
+ "relative_exclude_joints": [
154
+ "gripper"
155
+ ],
156
+ "optimizer_lr": 0.0001,
157
+ "optimizer_betas": [
158
+ 0.95,
159
+ 0.999
160
+ ],
161
+ "optimizer_eps": 1e-08,
162
+ "optimizer_weight_decay": 1e-05,
163
+ "warmup_ratio": 0.05,
164
+ "use_bf16": true,
165
+ "image_size": [
166
+ 256,
167
+ 256
168
+ ],
169
+ "tokenizer_assets_repo": null,
170
+ "lora_rank": 0,
171
+ "lora_alpha": 16,
172
+ "lora_dropout": 0.1,
173
+ "lora_full_model": false,
174
+ "video_backend": "decord",
175
+ "balance_dataset_weights": true,
176
+ "balance_trajectory_weights": true,
177
+ "dataset_paths": null,
178
+ "output_dir": "./tmp/gr00t",
179
+ "save_steps": 1000,
180
+ "max_steps": 10000,
181
+ "batch_size": 32,
182
+ "dataloader_num_workers": 8,
183
+ "report_to": "wandb",
184
+ "resume": false
185
+ },
186
+ "reward_model": null,
187
+ "output_dir": "outputs/train/sreetz-nv/so101-clean-up-vials-into-rack-50-groot-n17-relact-bs64-20k-20260628_lerobot_fix",
188
+ "job_name": "so101-clean-up-vials-into-rack-50-groot-n17-relact-bs64-20k-20260628_lerobot_fix",
189
+ "resume": true,
190
+ "seed": 42,
191
+ "cudnn_deterministic": false,
192
+ "num_workers": 4,
193
+ "batch_size": 64,
194
+ "prefetch_factor": 4,
195
+ "persistent_workers": true,
196
+ "steps": 20000,
197
+ "env_eval_freq": 0,
198
+ "log_freq": 100,
199
+ "eval_steps": 0,
200
+ "max_eval_samples": 0,
201
+ "tolerance_s": 0.0001,
202
+ "save_checkpoint": true,
203
+ "save_freq": 10000,
204
+ "use_policy_training_preset": true,
205
+ "optimizer": {
206
+ "type": "adamw",
207
+ "lr": 0.0001,
208
+ "weight_decay": 1e-05,
209
+ "grad_clip_norm": 10.0,
210
+ "betas": [
211
+ 0.95,
212
+ 0.999
213
+ ],
214
+ "eps": 1e-08
215
+ },
216
+ "scheduler": {
217
+ "type": "cosine_decay_with_warmup",
218
+ "num_warmup_steps": 500,
219
+ "num_decay_steps": 10000,
220
+ "peak_lr": 0.0001,
221
+ "decay_lr": 1e-05
222
+ },
223
+ "eval": {
224
+ "n_episodes": 50,
225
+ "batch_size": 22,
226
+ "use_async_envs": true,
227
+ "recording": false,
228
+ "recording_repo_id": null,
229
+ "recording_private": false
230
+ },
231
+ "wandb": {
232
+ "enable": false,
233
+ "disable_artifact": false,
234
+ "project": "lerobot",
235
+ "entity": null,
236
+ "notes": null,
237
+ "run_id": null,
238
+ "mode": null,
239
+ "add_tags": true
240
+ },
241
+ "peft": null,
242
+ "sample_weighting": null,
243
+ "rename_map": {},
244
+ "checkpoint_path": "outputs/train/sreetz-nv/so101-clean-up-vials-into-rack-50-groot-n17-relact-bs64-20k-20260628_lerobot_fix/checkpoints/last"
245
+ }