Robotics
LeRobot
Safetensors
groot
bimanual
yam
molmoact
helen9975 commited on
Commit
384ed60
·
verified ·
1 Parent(s): f995153

Upload fine-tuned checkpoint (groot-n1.7-molmoact-yam)

Browse files
README.md ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: lerobot
3
+ license: apache-2.0
4
+ tags:
5
+ - robotics
6
+ - lerobot
7
+ - groot
8
+ - bimanual
9
+ - yam
10
+ - molmoact
11
+ pipeline_tag: robotics
12
+ base_model: nvidia/GR00T-N1.7-3B
13
+ datasets:
14
+ - allenai/MolmoAct2-BimanualYAM-Dataset
15
+ ---
16
+
17
+ # GR00T N1.7 MolmoAct2 BimanualYAM
18
+
19
+ Partial fine-tune of [GR00T N1.7](https://huggingface.co/nvidia/GR00T-N1.7-3B) on the [MolmoAct2-BimanualYAM-Dataset](https://huggingface.co/datasets/allenai/MolmoAct2-BimanualYAM-Dataset) (3 cameras: top → left → right). Projector + diffusion trainable; LLM/vision frozen.
20
+
21
+ ## Checkpoint details
22
+
23
+ | Setting | Value |
24
+ |---------|-------|
25
+ | Training step | 30000 |
26
+ | Cameras | top, left, right |
27
+ | Action dim | 14 (absolute joints + grippers) |
28
+ | Chunk size | 16 |
29
+ | Embodiment | new_embodiment |
30
+ | Base model | nvidia/GR00T-N1.7-3B |
31
+ | Gripper range | [0, 1] (continuous) |
32
+
33
+ ## Usage (LeRobot)
34
+
35
+ ```python
36
+ from lerobot.policies.groot.modeling_groot import GrootPolicy
37
+ from lerobot.policies.factory import make_pre_post_processors
38
+
39
+ ckpt = "helen9975/groot-n1.7-molmoact-yam"
40
+ policy = GrootPolicy.from_pretrained(ckpt)
41
+ preprocessor, postprocessor = make_pre_post_processors(policy.config, pretrained_path=ckpt)
42
+ ```
43
+
44
+ ## Open-loop smoke (2 held-out episodes, stride 30, reset-each-step)
45
+
46
+ - **MSE:** 0.00160
47
+ - **MAE:** 0.0244
config.json ADDED
@@ -0,0 +1,109 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "groot",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.images.top": {
6
+ "type": "VISUAL",
7
+ "shape": [
8
+ 3,
9
+ 360,
10
+ 640
11
+ ]
12
+ },
13
+ "observation.images.left": {
14
+ "type": "VISUAL",
15
+ "shape": [
16
+ 3,
17
+ 360,
18
+ 640
19
+ ]
20
+ },
21
+ "observation.images.right": {
22
+ "type": "VISUAL",
23
+ "shape": [
24
+ 3,
25
+ 360,
26
+ 640
27
+ ]
28
+ },
29
+ "observation.state": {
30
+ "type": "STATE",
31
+ "shape": [
32
+ 14
33
+ ]
34
+ }
35
+ },
36
+ "output_features": {
37
+ "action": {
38
+ "type": "ACTION",
39
+ "shape": [
40
+ 14
41
+ ]
42
+ }
43
+ },
44
+ "device": "cuda",
45
+ "use_amp": false,
46
+ "use_peft": false,
47
+ "push_to_hub": false,
48
+ "repo_id": null,
49
+ "private": null,
50
+ "tags": null,
51
+ "license": null,
52
+ "pretrained_path": null,
53
+ "pretrained_revision": null,
54
+ "chunk_size": 16,
55
+ "n_action_steps": 16,
56
+ "max_state_dim": 132,
57
+ "max_action_dim": 132,
58
+ "normalization_mapping": {
59
+ "VISUAL": "IDENTITY",
60
+ "STATE": "IDENTITY",
61
+ "ACTION": "IDENTITY"
62
+ },
63
+ "base_model_path": "nvidia/GR00T-N1.7-3B",
64
+ "action_decode_transform": null,
65
+ "embodiment_tag": "new_embodiment",
66
+ "tune_llm": false,
67
+ "tune_visual": false,
68
+ "tune_projector": true,
69
+ "tune_diffusion_model": true,
70
+ "tune_vlln": true,
71
+ "tune_top_llm_layers": 0,
72
+ "num_inference_timesteps": null,
73
+ "rtc_ramp_rate": null,
74
+ "use_flash_attention": false,
75
+ "use_relative_actions": false,
76
+ "relative_exclude_joints": [
77
+ "gripper"
78
+ ],
79
+ "optimizer_lr": 0.0001,
80
+ "optimizer_betas": [
81
+ 0.9,
82
+ 0.999
83
+ ],
84
+ "optimizer_eps": 1e-08,
85
+ "optimizer_weight_decay": 1e-05,
86
+ "warmup_ratio": 0.05,
87
+ "use_bf16": true,
88
+ "model_params_fp32": true,
89
+ "image_size": [
90
+ 256,
91
+ 256
92
+ ],
93
+ "tokenizer_assets_repo": null,
94
+ "lora_rank": 0,
95
+ "lora_alpha": 16,
96
+ "lora_dropout": 0.1,
97
+ "lora_full_model": false,
98
+ "video_backend": "decord",
99
+ "balance_dataset_weights": true,
100
+ "balance_trajectory_weights": true,
101
+ "dataset_paths": null,
102
+ "output_dir": "./tmp/gr00t",
103
+ "save_steps": 1000,
104
+ "max_steps": 10000,
105
+ "batch_size": 32,
106
+ "dataloader_num_workers": 8,
107
+ "report_to": "wandb",
108
+ "resume": false
109
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:340ca7e9c99605ce41212fd0bc08cb258d2962d910ebfc34ee15564f3cd28d55
3
+ size 12576215296
policy_postprocessor.json ADDED
@@ -0,0 +1,23 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "policy_postprocessor",
3
+ "steps": [
4
+ {
5
+ "registry_name": "groot_action_unpack_unnormalize_v2",
6
+ "config": {
7
+ "env_action_dim": 14,
8
+ "normalize_min_max": true,
9
+ "clip_normalized_action": true,
10
+ "libero_gripper_action": false,
11
+ "libero_gripper_binarize": true
12
+ },
13
+ "state_file": "policy_postprocessor_step_0_groot_action_unpack_unnormalize_v2.safetensors"
14
+ },
15
+ {
16
+ "registry_name": "device_processor",
17
+ "config": {
18
+ "device": "cpu",
19
+ "float_dtype": null
20
+ }
21
+ }
22
+ ]
23
+ }
policy_postprocessor_step_0_groot_action_unpack_unnormalize_v2.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df8966df9dbfeb572de45955e84facf0946c6cd4f49f9ef7fa78c5a1fffab0f8
3
+ size 10720
policy_preprocessor.json ADDED
@@ -0,0 +1,78 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "policy_preprocessor",
3
+ "steps": [
4
+ {
5
+ "registry_name": "rename_observations_processor",
6
+ "config": {
7
+ "rename_map": {}
8
+ }
9
+ },
10
+ {
11
+ "registry_name": "to_batch_processor",
12
+ "config": {}
13
+ },
14
+ {
15
+ "registry_name": "groot_n1_7_pack_inputs_v1",
16
+ "config": {
17
+ "state_horizon": 1,
18
+ "action_horizon": 16,
19
+ "valid_action_horizon": 16,
20
+ "video_horizon": null,
21
+ "max_state_dim": 132,
22
+ "max_action_dim": 132,
23
+ "language_key": "task",
24
+ "formalize_language": true,
25
+ "embodiment_tag": "new_embodiment",
26
+ "embodiment_mapping": {
27
+ "oxe_droid_relative_eef_relative_joint": 24,
28
+ "xdof_relative_eef_relative_joint": 27,
29
+ "xdof_relative_eef_relative_joint_subtask": 27,
30
+ "real_g1_relative_eef_relative_joints": 25,
31
+ "real_r1_pro_sharpa_relative_eef": 26,
32
+ "real_r1_pro_sharpa_relative_eef_human": 26,
33
+ "real_r1_pro_sharpa_relative_eef_maxinsights": 26,
34
+ "real_r1_pro_sharpa_relative_eef_mecka": 26,
35
+ "unitree_g1_full_body_with_waist_height_nav_cmd": 25,
36
+ "simpler_env_google": 0,
37
+ "simpler_env_widowx": 1,
38
+ "libero_sim": 2,
39
+ "new_embodiment": 10
40
+ },
41
+ "normalize_min_max": true,
42
+ "state_dropout_prob": 0.0,
43
+ "clip_outliers": true,
44
+ "use_percentiles": false,
45
+ "video_modality_keys": null,
46
+ "raw_stats": null,
47
+ "modality_config": null
48
+ },
49
+ "state_file": "policy_preprocessor_step_2_groot_n1_7_pack_inputs_v1.safetensors"
50
+ },
51
+ {
52
+ "registry_name": "groot_n1_7_vlm_encode_v1",
53
+ "config": {
54
+ "model_name": "Qwen/Qwen3-VL-2B-Instruct",
55
+ "image_crop_size": [
56
+ 230,
57
+ 230
58
+ ],
59
+ "image_target_size": [
60
+ 256,
61
+ 256
62
+ ],
63
+ "shortest_image_edge": null,
64
+ "crop_fraction": null,
65
+ "use_albumentations": false,
66
+ "letter_box_transform": false,
67
+ "device": "cuda"
68
+ }
69
+ },
70
+ {
71
+ "registry_name": "device_processor",
72
+ "config": {
73
+ "device": "cuda",
74
+ "float_dtype": null
75
+ }
76
+ }
77
+ ]
78
+ }
policy_preprocessor_step_2_groot_n1_7_pack_inputs_v1.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df8966df9dbfeb572de45955e84facf0946c6cd4f49f9ef7fa78c5a1fffab0f8
3
+ size 10720
train_config.json ADDED
@@ -0,0 +1,260 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "allenai/MolmoAct2-BimanualYAM-Dataset",
4
+ "root": "/weka/oe-training-default/molmoact/lerobot/allenai/MolmoAct2-BimanualYAM-Dataset",
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": "local",
78
+ "use_imagenet_stats": true,
79
+ "video_backend": "pyav",
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.images.top": {
91
+ "type": "VISUAL",
92
+ "shape": [
93
+ 3,
94
+ 360,
95
+ 640
96
+ ]
97
+ },
98
+ "observation.images.left": {
99
+ "type": "VISUAL",
100
+ "shape": [
101
+ 3,
102
+ 360,
103
+ 640
104
+ ]
105
+ },
106
+ "observation.images.right": {
107
+ "type": "VISUAL",
108
+ "shape": [
109
+ 3,
110
+ 360,
111
+ 640
112
+ ]
113
+ },
114
+ "observation.state": {
115
+ "type": "STATE",
116
+ "shape": [
117
+ 14
118
+ ]
119
+ }
120
+ },
121
+ "output_features": {
122
+ "action": {
123
+ "type": "ACTION",
124
+ "shape": [
125
+ 14
126
+ ]
127
+ }
128
+ },
129
+ "device": "cuda",
130
+ "use_amp": false,
131
+ "use_peft": false,
132
+ "push_to_hub": false,
133
+ "repo_id": null,
134
+ "private": null,
135
+ "tags": null,
136
+ "license": null,
137
+ "pretrained_path": null,
138
+ "pretrained_revision": null,
139
+ "chunk_size": 16,
140
+ "n_action_steps": 16,
141
+ "max_state_dim": 132,
142
+ "max_action_dim": 132,
143
+ "normalization_mapping": {
144
+ "VISUAL": "IDENTITY",
145
+ "STATE": "IDENTITY",
146
+ "ACTION": "IDENTITY"
147
+ },
148
+ "base_model_path": "nvidia/GR00T-N1.7-3B",
149
+ "action_decode_transform": null,
150
+ "embodiment_tag": "new_embodiment",
151
+ "tune_llm": false,
152
+ "tune_visual": false,
153
+ "tune_projector": true,
154
+ "tune_diffusion_model": true,
155
+ "tune_vlln": true,
156
+ "tune_top_llm_layers": 0,
157
+ "num_inference_timesteps": null,
158
+ "rtc_ramp_rate": null,
159
+ "use_flash_attention": false,
160
+ "use_relative_actions": false,
161
+ "relative_exclude_joints": [
162
+ "gripper"
163
+ ],
164
+ "optimizer_lr": 0.0001,
165
+ "optimizer_betas": [
166
+ 0.9,
167
+ 0.999
168
+ ],
169
+ "optimizer_eps": 1e-08,
170
+ "optimizer_weight_decay": 1e-05,
171
+ "warmup_ratio": 0.05,
172
+ "use_bf16": true,
173
+ "model_params_fp32": true,
174
+ "image_size": [
175
+ 256,
176
+ 256
177
+ ],
178
+ "tokenizer_assets_repo": null,
179
+ "lora_rank": 0,
180
+ "lora_alpha": 16,
181
+ "lora_dropout": 0.1,
182
+ "lora_full_model": false,
183
+ "video_backend": "decord",
184
+ "balance_dataset_weights": true,
185
+ "balance_trajectory_weights": true,
186
+ "dataset_paths": null,
187
+ "output_dir": "./tmp/gr00t",
188
+ "save_steps": 1000,
189
+ "max_steps": 10000,
190
+ "batch_size": 32,
191
+ "dataloader_num_workers": 8,
192
+ "report_to": "wandb",
193
+ "resume": false
194
+ },
195
+ "reward_model": null,
196
+ "output_dir": "/weka/robots-default/helenw/outputs/groot_1",
197
+ "job_name": "groot_1",
198
+ "resume": false,
199
+ "seed": 1000,
200
+ "cudnn_deterministic": false,
201
+ "num_workers": 4,
202
+ "batch_size": 128,
203
+ "prefetch_factor": 2,
204
+ "persistent_workers": true,
205
+ "steps": 30000,
206
+ "env_eval_freq": 0,
207
+ "log_freq": 10,
208
+ "eval_steps": 0,
209
+ "max_eval_samples": 0,
210
+ "tolerance_s": 0.04,
211
+ "save_checkpoint": true,
212
+ "save_freq": 5000,
213
+ "use_policy_training_preset": true,
214
+ "optimizer": {
215
+ "type": "adamw",
216
+ "lr": 0.0001,
217
+ "weight_decay": 1e-05,
218
+ "grad_clip_norm": 1.0,
219
+ "betas": [
220
+ 0.9,
221
+ 0.999
222
+ ],
223
+ "eps": 1e-08
224
+ },
225
+ "scheduler": {
226
+ "type": "diffuser",
227
+ "num_warmup_steps": 500,
228
+ "name": "cosine"
229
+ },
230
+ "eval": {
231
+ "n_episodes": 50,
232
+ "batch_size": 50,
233
+ "use_async_envs": true,
234
+ "recording": false,
235
+ "recording_repo_id": null,
236
+ "recording_private": false
237
+ },
238
+ "wandb": {
239
+ "enable": true,
240
+ "disable_artifact": true,
241
+ "project": "lerobot",
242
+ "entity": null,
243
+ "notes": null,
244
+ "run_id": "d7bs1z5z",
245
+ "mode": null,
246
+ "add_tags": true
247
+ },
248
+ "peft": null,
249
+ "job": {
250
+ "target": null,
251
+ "image": "huggingface/lerobot-gpu:latest",
252
+ "timeout": "2d",
253
+ "detach": false,
254
+ "tags": []
255
+ },
256
+ "save_checkpoint_to_hub": false,
257
+ "sample_weighting": null,
258
+ "rename_map": {},
259
+ "checkpoint_path": null
260
+ }