Robotics
LeRobot
Safetensors
pi05
rrrfy commited on
Commit
cc9219c
·
verified ·
1 Parent(s): 19e82e6

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +70 -0
  2. config.json +139 -0
  3. model.safetensors +3 -0
  4. train_config.json +277 -0
README.md ADDED
@@ -0,0 +1,70 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: unitreerobotics/G1_Dex3_ObjectPlacement_Dataset
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: pi05
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - lerobot
9
+ - robotics
10
+ - pi05
11
+ ---
12
+
13
+ # Model Card for pi05
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+ **π₀.₅ (Pi05) Policy**
19
+
20
+ π₀.₅ is a Vision-Language-Action model with open-world generalization, from Physical Intelligence. The LeRobot implementation is adapted from their open source OpenPI repository.
21
+
22
+ **Model Overview**
23
+
24
+ π₀.₅ represents a significant evolution from π₀, developed by Physical Intelligence to address a big challenge in robotics: open-world generalization. While robots can perform impressive tasks in controlled environments, π₀.₅ is designed to generalize to entirely new environments and situations that were never seen during training.
25
+
26
+ For more details, see the [Physical Intelligence π₀.₅ blog post](https://www.physicalintelligence.company/blog/pi05).
27
+
28
+
29
+ This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
30
+ See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).
31
+
32
+ ---
33
+
34
+ ## How to Get Started with the Model
35
+
36
+ For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
37
+ Below is the short version on how to train and run inference/eval:
38
+
39
+ ### Train from scratch
40
+
41
+ ```bash
42
+ lerobot-train \
43
+ --dataset.repo_id=${HF_USER}/<dataset> \
44
+ --policy.type=act \
45
+ --output_dir=outputs/train/<desired_policy_repo_id> \
46
+ --job_name=lerobot_training \
47
+ --policy.device=cuda \
48
+ --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
49
+ --wandb.enable=true
50
+ ```
51
+
52
+ _Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`._
53
+
54
+ ### Evaluate the policy/run inference
55
+
56
+ ```bash
57
+ lerobot-record \
58
+ --robot.type=so100_follower \
59
+ --dataset.repo_id=<hf_user>/eval_<dataset> \
60
+ --policy.path=<hf_user>/<desired_policy_repo_id> \
61
+ --episodes=10
62
+ ```
63
+
64
+ Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.
65
+
66
+ ---
67
+
68
+ ## Model Details
69
+
70
+ - **License:** apache-2.0
config.json ADDED
@@ -0,0 +1,139 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "pi05",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 28
9
+ ]
10
+ },
11
+ "observation.images.cam_left_high": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 480,
16
+ 640
17
+ ]
18
+ },
19
+ "observation.images.cam_right_high": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 480,
24
+ 640
25
+ ]
26
+ },
27
+ "observation.images.cam_left_wrist": {
28
+ "type": "VISUAL",
29
+ "shape": [
30
+ 3,
31
+ 480,
32
+ 640
33
+ ]
34
+ },
35
+ "observation.images.cam_right_wrist": {
36
+ "type": "VISUAL",
37
+ "shape": [
38
+ 3,
39
+ 480,
40
+ 640
41
+ ]
42
+ }
43
+ },
44
+ "output_features": {
45
+ "action": {
46
+ "type": "ACTION",
47
+ "shape": [
48
+ 28
49
+ ]
50
+ }
51
+ },
52
+ "device": "cuda",
53
+ "use_amp": false,
54
+ "use_peft": false,
55
+ "push_to_hub": true,
56
+ "repo_id": "rrrfy/pi05",
57
+ "private": null,
58
+ "tags": null,
59
+ "license": null,
60
+ "pretrained_path": "lerobot/pi05_base",
61
+ "paligemma_variant": "gemma_2b",
62
+ "action_expert_variant": "gemma_300m",
63
+ "dtype": "bfloat16",
64
+ "chunk_size": 50,
65
+ "n_action_steps": 50,
66
+ "max_state_dim": 32,
67
+ "max_action_dim": 32,
68
+ "num_inference_steps": 10,
69
+ "time_sampling_beta_alpha": 1.5,
70
+ "time_sampling_beta_beta": 1.0,
71
+ "time_sampling_scale": 0.999,
72
+ "time_sampling_offset": 0.001,
73
+ "min_period": 0.004,
74
+ "max_period": 4.0,
75
+ "use_relative_actions": false,
76
+ "relative_exclude_joints": [
77
+ "gripper"
78
+ ],
79
+ "action_feature_names": [
80
+ [
81
+ "kLeftShoulderPitch",
82
+ "kLeftShoulderRoll",
83
+ "kLeftShoulderYaw",
84
+ "kLeftElbow",
85
+ "kLeftWristRoll",
86
+ "kLeftWristPitch",
87
+ "kLeftWristYaw",
88
+ "kRightShoulderPitch",
89
+ "kRightShoulderRoll",
90
+ "kRightShoulderYaw",
91
+ "kRightElbow",
92
+ "kRightWristRoll",
93
+ "kRightWristPitch",
94
+ "kRightWristYaw",
95
+ "kLeftHandThumb0",
96
+ "kLeftHandThumb1",
97
+ "kLeftHandThumb2",
98
+ "kLeftHandMiddle0",
99
+ "kLeftHandMiddle1",
100
+ "kLeftHandIndex0",
101
+ "kLeftHandIndex1",
102
+ "kRightHandThumb0",
103
+ "kRightHandThumb1",
104
+ "kRightHandThumb2",
105
+ "kRightHandIndex0",
106
+ "kRightHandIndex1",
107
+ "kRightHandMiddle0",
108
+ "kRightHandMiddle1"
109
+ ]
110
+ ],
111
+ "rtc_config": null,
112
+ "image_resolution": [
113
+ 224,
114
+ 224
115
+ ],
116
+ "empty_cameras": 0,
117
+ "tokenizer_max_length": 200,
118
+ "normalization_mapping": {
119
+ "VISUAL": "IDENTITY",
120
+ "STATE": "QUANTILES",
121
+ "ACTION": "QUANTILES"
122
+ },
123
+ "gradient_checkpointing": true,
124
+ "compile_model": false,
125
+ "compile_mode": "max-autotune",
126
+ "freeze_vision_encoder": false,
127
+ "train_expert_only": true,
128
+ "optimizer_lr": 2.5e-05,
129
+ "optimizer_betas": [
130
+ 0.9,
131
+ 0.95
132
+ ],
133
+ "optimizer_eps": 1e-08,
134
+ "optimizer_weight_decay": 0.01,
135
+ "optimizer_grad_clip_norm": 1.0,
136
+ "scheduler_warmup_steps": 1000,
137
+ "scheduler_decay_steps": 30000,
138
+ "scheduler_decay_lr": 2.5e-06
139
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f4bcc3a56362f17217ba00b1847d727620eabe4b2ccd3747780089aeedaaba4
3
+ size 9354050752
train_config.json ADDED
@@ -0,0 +1,277 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "unitreerobotics/G1_Dex3_ObjectPlacement_Dataset",
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
+ "return_uint8": false,
81
+ "streaming": false
82
+ },
83
+ "env": null,
84
+ "policy": {
85
+ "type": "pi05",
86
+ "n_obs_steps": 1,
87
+ "input_features": {
88
+ "observation.state": {
89
+ "type": "STATE",
90
+ "shape": [
91
+ 28
92
+ ]
93
+ },
94
+ "observation.images.cam_left_high": {
95
+ "type": "VISUAL",
96
+ "shape": [
97
+ 3,
98
+ 480,
99
+ 640
100
+ ]
101
+ },
102
+ "observation.images.cam_right_high": {
103
+ "type": "VISUAL",
104
+ "shape": [
105
+ 3,
106
+ 480,
107
+ 640
108
+ ]
109
+ },
110
+ "observation.images.cam_left_wrist": {
111
+ "type": "VISUAL",
112
+ "shape": [
113
+ 3,
114
+ 480,
115
+ 640
116
+ ]
117
+ },
118
+ "observation.images.cam_right_wrist": {
119
+ "type": "VISUAL",
120
+ "shape": [
121
+ 3,
122
+ 480,
123
+ 640
124
+ ]
125
+ }
126
+ },
127
+ "output_features": {
128
+ "action": {
129
+ "type": "ACTION",
130
+ "shape": [
131
+ 28
132
+ ]
133
+ }
134
+ },
135
+ "device": "cuda",
136
+ "use_amp": false,
137
+ "use_peft": false,
138
+ "push_to_hub": true,
139
+ "repo_id": "rrrfy/pi05",
140
+ "private": null,
141
+ "tags": null,
142
+ "license": null,
143
+ "pretrained_path": "lerobot/pi05_base",
144
+ "paligemma_variant": "gemma_2b",
145
+ "action_expert_variant": "gemma_300m",
146
+ "dtype": "bfloat16",
147
+ "chunk_size": 50,
148
+ "n_action_steps": 50,
149
+ "max_state_dim": 32,
150
+ "max_action_dim": 32,
151
+ "num_inference_steps": 10,
152
+ "time_sampling_beta_alpha": 1.5,
153
+ "time_sampling_beta_beta": 1.0,
154
+ "time_sampling_scale": 0.999,
155
+ "time_sampling_offset": 0.001,
156
+ "min_period": 0.004,
157
+ "max_period": 4.0,
158
+ "use_relative_actions": false,
159
+ "relative_exclude_joints": [
160
+ "gripper"
161
+ ],
162
+ "action_feature_names": [
163
+ [
164
+ "kLeftShoulderPitch",
165
+ "kLeftShoulderRoll",
166
+ "kLeftShoulderYaw",
167
+ "kLeftElbow",
168
+ "kLeftWristRoll",
169
+ "kLeftWristPitch",
170
+ "kLeftWristYaw",
171
+ "kRightShoulderPitch",
172
+ "kRightShoulderRoll",
173
+ "kRightShoulderYaw",
174
+ "kRightElbow",
175
+ "kRightWristRoll",
176
+ "kRightWristPitch",
177
+ "kRightWristYaw",
178
+ "kLeftHandThumb0",
179
+ "kLeftHandThumb1",
180
+ "kLeftHandThumb2",
181
+ "kLeftHandMiddle0",
182
+ "kLeftHandMiddle1",
183
+ "kLeftHandIndex0",
184
+ "kLeftHandIndex1",
185
+ "kRightHandThumb0",
186
+ "kRightHandThumb1",
187
+ "kRightHandThumb2",
188
+ "kRightHandIndex0",
189
+ "kRightHandIndex1",
190
+ "kRightHandMiddle0",
191
+ "kRightHandMiddle1"
192
+ ]
193
+ ],
194
+ "rtc_config": null,
195
+ "image_resolution": [
196
+ 224,
197
+ 224
198
+ ],
199
+ "empty_cameras": 0,
200
+ "tokenizer_max_length": 200,
201
+ "normalization_mapping": {
202
+ "VISUAL": "IDENTITY",
203
+ "STATE": "QUANTILES",
204
+ "ACTION": "QUANTILES"
205
+ },
206
+ "gradient_checkpointing": true,
207
+ "compile_model": false,
208
+ "compile_mode": "max-autotune",
209
+ "freeze_vision_encoder": false,
210
+ "train_expert_only": true,
211
+ "optimizer_lr": 2.5e-05,
212
+ "optimizer_betas": [
213
+ 0.9,
214
+ 0.95
215
+ ],
216
+ "optimizer_eps": 1e-08,
217
+ "optimizer_weight_decay": 0.01,
218
+ "optimizer_grad_clip_norm": 1.0,
219
+ "scheduler_warmup_steps": 1000,
220
+ "scheduler_decay_steps": 30000,
221
+ "scheduler_decay_lr": 2.5e-06
222
+ },
223
+ "reward_model": null,
224
+ "output_dir": "outputs/pi05_training",
225
+ "job_name": "pi05_training",
226
+ "resume": false,
227
+ "seed": 1000,
228
+ "cudnn_deterministic": false,
229
+ "num_workers": 4,
230
+ "batch_size": 8,
231
+ "prefetch_factor": 4,
232
+ "persistent_workers": true,
233
+ "steps": 3000,
234
+ "eval_freq": 20000,
235
+ "log_freq": 200,
236
+ "tolerance_s": 0.0001,
237
+ "save_checkpoint": true,
238
+ "save_freq": 20000,
239
+ "use_policy_training_preset": true,
240
+ "optimizer": {
241
+ "type": "adamw",
242
+ "lr": 2.5e-05,
243
+ "weight_decay": 0.01,
244
+ "grad_clip_norm": 1.0,
245
+ "betas": [
246
+ 0.9,
247
+ 0.95
248
+ ],
249
+ "eps": 1e-08
250
+ },
251
+ "scheduler": {
252
+ "type": "cosine_decay_with_warmup",
253
+ "num_warmup_steps": 1000,
254
+ "num_decay_steps": 30000,
255
+ "peak_lr": 2.5e-05,
256
+ "decay_lr": 2.5e-06
257
+ },
258
+ "eval": {
259
+ "n_episodes": 50,
260
+ "batch_size": 22,
261
+ "use_async_envs": true
262
+ },
263
+ "wandb": {
264
+ "enable": false,
265
+ "disable_artifact": false,
266
+ "project": "lerobot",
267
+ "entity": null,
268
+ "notes": null,
269
+ "run_id": null,
270
+ "mode": null,
271
+ "add_tags": true
272
+ },
273
+ "peft": null,
274
+ "sample_weighting": null,
275
+ "rename_map": {},
276
+ "checkpoint_path": null
277
+ }