Hugging Face's logo

XYZPIT
/
260512_r1lite_diffusion

Robotics
LeRobot
Safetensors
diffusion
Model card Files
xet
 Community
woozziam commited on
Commit
5d3a14d
·
verified ·
1 Parent(s): 1ebc6ab

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +163 -0
  3. model.safetensors +3 -0
  4. train_config.json +302 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: woozziam/260413_r1lite_lerobot
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: diffusion
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - lerobot
9
+ - robotics
10
+ - diffusion
11
+ ---
12
+
13
+ # Model Card for diffusion
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+ [Diffusion Policy](https://huggingface.co/papers/2303.04137) treats visuomotor control as a generative diffusion process, producing smooth, multi-step action trajectories that excel at contact-rich manipulation.
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 +1,163 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "diffusion",
3
+ "n_obs_steps": 2,
4
+ "input_features": {
5
+ "observation.images.head_rgb": {
6
+ "type": "VISUAL",
7
+ "shape": [
8
+ 3,
9
+ 96,
10
+ 96
11
+ ]
12
+ },
13
+ "observation.images.left_wrist_rgb": {
14
+ "type": "VISUAL",
15
+ "shape": [
16
+ 3,
17
+ 96,
18
+ 96
19
+ ]
20
+ },
21
+ "observation.images.right_wrist_rgb": {
22
+ "type": "VISUAL",
23
+ "shape": [
24
+ 3,
25
+ 96,
26
+ 96
27
+ ]
28
+ },
29
+ "observation.state.left_arm": {
30
+ "type": "STATE",
31
+ "shape": [
32
+ 6
33
+ ]
34
+ },
35
+ "observation.state.right_arm": {
36
+ "type": "STATE",
37
+ "shape": [
38
+ 6
39
+ ]
40
+ },
41
+ "observation.state.left_gripper": {
42
+ "type": "STATE",
43
+ "shape": [
44
+ 1
45
+ ]
46
+ },
47
+ "observation.state.right_gripper": {
48
+ "type": "STATE",
49
+ "shape": [
50
+ 1
51
+ ]
52
+ },
53
+ "observation.state": {
54
+ "type": "STATE",
55
+ "shape": [
56
+ 14
57
+ ]
58
+ }
59
+ },
60
+ "output_features": {
61
+ "action.left_gripper": {
62
+ "type": "ACTION",
63
+ "shape": [
64
+ 1
65
+ ]
66
+ },
67
+ "action.right_gripper": {
68
+ "type": "ACTION",
69
+ "shape": [
70
+ 1
71
+ ]
72
+ },
73
+ "action.chassis.velocities": {
74
+ "type": "ACTION",
75
+ "shape": [
76
+ 6
77
+ ]
78
+ },
79
+ "action.torso.velocities": {
80
+ "type": "ACTION",
81
+ "shape": [
82
+ 6
83
+ ]
84
+ },
85
+ "action.left_arm": {
86
+ "type": "ACTION",
87
+ "shape": [
88
+ 6
89
+ ]
90
+ },
91
+ "action.right_arm": {
92
+ "type": "ACTION",
93
+ "shape": [
94
+ 6
95
+ ]
96
+ },
97
+ "action": {
98
+ "type": "ACTION",
99
+ "shape": [
100
+ 14
101
+ ]
102
+ }
103
+ },
104
+ "device": "cuda",
105
+ "use_amp": false,
106
+ "use_peft": false,
107
+ "push_to_hub": true,
108
+ "repo_id": "woozziam/260512_r1lite_diffusion",
109
+ "private": null,
110
+ "tags": null,
111
+ "license": null,
112
+ "pretrained_path": null,
113
+ "horizon": 32,
114
+ "n_action_steps": 30,
115
+ "normalization_mapping": {
116
+ "VISUAL": "MEAN_STD",
117
+ "STATE": "MIN_MAX",
118
+ "ACTION": "MIN_MAX"
119
+ },
120
+ "drop_n_last_frames": 1,
121
+ "vision_backbone": "resnet18",
122
+ "resize_shape": [
123
+ 96,
124
+ 96
125
+ ],
126
+ "crop_ratio": 1.0,
127
+ "crop_shape": null,
128
+ "crop_is_random": true,
129
+ "pretrained_backbone_weights": null,
130
+ "use_group_norm": true,
131
+ "spatial_softmax_num_keypoints": 32,
132
+ "use_separate_rgb_encoder_per_camera": false,
133
+ "down_dims": [
134
+ 512,
135
+ 1024,
136
+ 2048
137
+ ],
138
+ "kernel_size": 5,
139
+ "n_groups": 8,
140
+ "diffusion_step_embed_dim": 128,
141
+ "use_film_scale_modulation": true,
142
+ "noise_scheduler_type": "DDIM",
143
+ "num_train_timesteps": 100,
144
+ "beta_schedule": "squaredcos_cap_v2",
145
+ "beta_start": 0.0001,
146
+ "beta_end": 0.02,
147
+ "prediction_type": "epsilon",
148
+ "clip_sample": true,
149
+ "clip_sample_range": 1.0,
150
+ "num_inference_steps": 20,
151
+ "compile_model": false,
152
+ "compile_mode": "reduce-overhead",
153
+ "do_mask_loss_for_padding": false,
154
+ "optimizer_lr": 0.0001,
155
+ "optimizer_betas": [
156
+ 0.95,
157
+ 0.999
158
+ ],
159
+ "optimizer_eps": 1e-08,
160
+ "optimizer_weight_decay": 1e-06,
161
+ "scheduler_name": "cosine",
162
+ "scheduler_warmup_steps": 500
163
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70e459750f6f101e3f5758ac413ee09137548970187ea3904eb5551d9e42ce37
3
+ size 1083146184
train_config.json ADDED
@@ -0,0 +1,302 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "woozziam/260413_r1lite_lerobot",
4
+ "root": "/home/wook/lerobot/woozziam/260413_r1lite_lerobot",
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
+ "streaming": false
82
+ },
83
+ "env": null,
84
+ "policy": {
85
+ "type": "diffusion",
86
+ "n_obs_steps": 2,
87
+ "input_features": {
88
+ "observation.images.head_rgb": {
89
+ "type": "VISUAL",
90
+ "shape": [
91
+ 3,
92
+ 96,
93
+ 96
94
+ ]
95
+ },
96
+ "observation.images.left_wrist_rgb": {
97
+ "type": "VISUAL",
98
+ "shape": [
99
+ 3,
100
+ 96,
101
+ 96
102
+ ]
103
+ },
104
+ "observation.images.right_wrist_rgb": {
105
+ "type": "VISUAL",
106
+ "shape": [
107
+ 3,
108
+ 96,
109
+ 96
110
+ ]
111
+ },
112
+ "observation.state.left_arm": {
113
+ "type": "STATE",
114
+ "shape": [
115
+ 6
116
+ ]
117
+ },
118
+ "observation.state.right_arm": {
119
+ "type": "STATE",
120
+ "shape": [
121
+ 6
122
+ ]
123
+ },
124
+ "observation.state.left_gripper": {
125
+ "type": "STATE",
126
+ "shape": [
127
+ 1
128
+ ]
129
+ },
130
+ "observation.state.right_gripper": {
131
+ "type": "STATE",
132
+ "shape": [
133
+ 1
134
+ ]
135
+ },
136
+ "observation.state": {
137
+ "type": "STATE",
138
+ "shape": [
139
+ 14
140
+ ]
141
+ }
142
+ },
143
+ "output_features": {
144
+ "action.left_gripper": {
145
+ "type": "ACTION",
146
+ "shape": [
147
+ 1
148
+ ]
149
+ },
150
+ "action.right_gripper": {
151
+ "type": "ACTION",
152
+ "shape": [
153
+ 1
154
+ ]
155
+ },
156
+ "action.chassis.velocities": {
157
+ "type": "ACTION",
158
+ "shape": [
159
+ 6
160
+ ]
161
+ },
162
+ "action.torso.velocities": {
163
+ "type": "ACTION",
164
+ "shape": [
165
+ 6
166
+ ]
167
+ },
168
+ "action.left_arm": {
169
+ "type": "ACTION",
170
+ "shape": [
171
+ 6
172
+ ]
173
+ },
174
+ "action.right_arm": {
175
+ "type": "ACTION",
176
+ "shape": [
177
+ 6
178
+ ]
179
+ },
180
+ "action": {
181
+ "type": "ACTION",
182
+ "shape": [
183
+ 14
184
+ ]
185
+ }
186
+ },
187
+ "device": "cuda",
188
+ "use_amp": false,
189
+ "use_peft": false,
190
+ "push_to_hub": true,
191
+ "repo_id": "woozziam/260512_r1lite_diffusion",
192
+ "private": null,
193
+ "tags": null,
194
+ "license": null,
195
+ "pretrained_path": null,
196
+ "horizon": 32,
197
+ "n_action_steps": 30,
198
+ "normalization_mapping": {
199
+ "VISUAL": "MEAN_STD",
200
+ "STATE": "MIN_MAX",
201
+ "ACTION": "MIN_MAX"
202
+ },
203
+ "drop_n_last_frames": 1,
204
+ "vision_backbone": "resnet18",
205
+ "resize_shape": [
206
+ 96,
207
+ 96
208
+ ],
209
+ "crop_ratio": 1.0,
210
+ "crop_shape": null,
211
+ "crop_is_random": true,
212
+ "pretrained_backbone_weights": null,
213
+ "use_group_norm": true,
214
+ "spatial_softmax_num_keypoints": 32,
215
+ "use_separate_rgb_encoder_per_camera": false,
216
+ "down_dims": [
217
+ 512,
218
+ 1024,
219
+ 2048
220
+ ],
221
+ "kernel_size": 5,
222
+ "n_groups": 8,
223
+ "diffusion_step_embed_dim": 128,
224
+ "use_film_scale_modulation": true,
225
+ "noise_scheduler_type": "DDIM",
226
+ "num_train_timesteps": 100,
227
+ "beta_schedule": "squaredcos_cap_v2",
228
+ "beta_start": 0.0001,
229
+ "beta_end": 0.02,
230
+ "prediction_type": "epsilon",
231
+ "clip_sample": true,
232
+ "clip_sample_range": 1.0,
233
+ "num_inference_steps": 20,
234
+ "compile_model": false,
235
+ "compile_mode": "reduce-overhead",
236
+ "do_mask_loss_for_padding": false,
237
+ "optimizer_lr": 0.0001,
238
+ "optimizer_betas": [
239
+ 0.95,
240
+ 0.999
241
+ ],
242
+ "optimizer_eps": 1e-08,
243
+ "optimizer_weight_decay": 1e-06,
244
+ "scheduler_name": "cosine",
245
+ "scheduler_warmup_steps": 500
246
+ },
247
+ "output_dir": "outputs/train/260512_r1lite_diffusion",
248
+ "job_name": "260512_r1lite_diffusion",
249
+ "resume": false,
250
+ "seed": 1000,
251
+ "cudnn_deterministic": false,
252
+ "num_workers": 4,
253
+ "batch_size": 8,
254
+ "prefetch_factor": 4,
255
+ "persistent_workers": true,
256
+ "steps": 200000,
257
+ "eval_freq": 20000,
258
+ "log_freq": 500,
259
+ "tolerance_s": 0.0001,
260
+ "save_checkpoint": true,
261
+ "save_freq": 20000,
262
+ "use_policy_training_preset": true,
263
+ "optimizer": {
264
+ "type": "adam",
265
+ "lr": 0.0001,
266
+ "weight_decay": 1e-06,
267
+ "grad_clip_norm": 10.0,
268
+ "betas": [
269
+ 0.95,
270
+ 0.999
271
+ ],
272
+ "eps": 1e-08
273
+ },
274
+ "scheduler": {
275
+ "type": "diffuser",
276
+ "num_warmup_steps": 500,
277
+ "name": "cosine"
278
+ },
279
+ "eval": {
280
+ "n_episodes": 50,
281
+ "batch_size": 16,
282
+ "use_async_envs": true
283
+ },
284
+ "wandb": {
285
+ "enable": true,
286
+ "disable_artifact": false,
287
+ "project": "lerobot",
288
+ "entity": null,
289
+ "notes": null,
290
+ "run_id": "7v4yzr1u",
291
+ "mode": null,
292
+ "add_tags": true
293
+ },
294
+ "peft": null,
295
+ "use_rabc": false,
296
+ "rabc_progress_path": null,
297
+ "rabc_kappa": 0.01,
298
+ "rabc_epsilon": 1e-06,
299
+ "rabc_head_mode": "sparse",
300
+ "rename_map": {},
301
+ "checkpoint_path": null
302
+ }