ewayWu commited on
Commit
81a3e59
·
verified ·
1 Parent(s): a67ab04

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +99 -0
  3. model.safetensors +3 -0
  4. train_config.json +233 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: local/woosh_pickup_takeout
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: diffusion
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - diffusion
9
+ - robotics
10
+ - lerobot
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 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "diffusion",
3
+ "n_obs_steps": 2,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 38
9
+ ]
10
+ },
11
+ "observation.images.head": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 480,
16
+ 848
17
+ ]
18
+ },
19
+ "observation.images.left_wrist": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 480,
24
+ 848
25
+ ]
26
+ },
27
+ "observation.images.right_wrist": {
28
+ "type": "VISUAL",
29
+ "shape": [
30
+ 3,
31
+ 480,
32
+ 848
33
+ ]
34
+ }
35
+ },
36
+ "output_features": {
37
+ "action": {
38
+ "type": "ACTION",
39
+ "shape": [
40
+ 27
41
+ ]
42
+ }
43
+ },
44
+ "device": "cuda",
45
+ "use_amp": false,
46
+ "use_peft": false,
47
+ "push_to_hub": true,
48
+ "repo_id": "ewayWu/diffusion_policy_pickup_takeout",
49
+ "private": null,
50
+ "tags": null,
51
+ "license": null,
52
+ "pretrained_path": null,
53
+ "horizon": 16,
54
+ "n_action_steps": 8,
55
+ "normalization_mapping": {
56
+ "VISUAL": "MEAN_STD",
57
+ "STATE": "MIN_MAX",
58
+ "ACTION": "MIN_MAX"
59
+ },
60
+ "drop_n_last_frames": 7,
61
+ "vision_backbone": "resnet18",
62
+ "crop_shape": [
63
+ 84,
64
+ 84
65
+ ],
66
+ "crop_is_random": true,
67
+ "pretrained_backbone_weights": null,
68
+ "use_group_norm": true,
69
+ "spatial_softmax_num_keypoints": 32,
70
+ "use_separate_rgb_encoder_per_camera": false,
71
+ "down_dims": [
72
+ 512,
73
+ 1024,
74
+ 2048
75
+ ],
76
+ "kernel_size": 5,
77
+ "n_groups": 8,
78
+ "diffusion_step_embed_dim": 128,
79
+ "use_film_scale_modulation": true,
80
+ "noise_scheduler_type": "DDPM",
81
+ "num_train_timesteps": 100,
82
+ "beta_schedule": "squaredcos_cap_v2",
83
+ "beta_start": 0.0001,
84
+ "beta_end": 0.02,
85
+ "prediction_type": "epsilon",
86
+ "clip_sample": true,
87
+ "clip_sample_range": 1.0,
88
+ "num_inference_steps": null,
89
+ "do_mask_loss_for_padding": false,
90
+ "optimizer_lr": 0.0001,
91
+ "optimizer_betas": [
92
+ 0.95,
93
+ 0.999
94
+ ],
95
+ "optimizer_eps": 1e-08,
96
+ "optimizer_weight_decay": 1e-06,
97
+ "scheduler_name": "cosine",
98
+ "scheduler_warmup_steps": 500
99
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f916387393b43a987f362ef3c92a1a253ebd21b674a2572889fe60aea1ffccb1
3
+ size 1088837636
train_config.json ADDED
@@ -0,0 +1,233 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "local/woosh_pickup_takeout",
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": "diffusion",
85
+ "n_obs_steps": 2,
86
+ "input_features": {
87
+ "observation.state": {
88
+ "type": "STATE",
89
+ "shape": [
90
+ 38
91
+ ]
92
+ },
93
+ "observation.images.head": {
94
+ "type": "VISUAL",
95
+ "shape": [
96
+ 3,
97
+ 480,
98
+ 848
99
+ ]
100
+ },
101
+ "observation.images.left_wrist": {
102
+ "type": "VISUAL",
103
+ "shape": [
104
+ 3,
105
+ 480,
106
+ 848
107
+ ]
108
+ },
109
+ "observation.images.right_wrist": {
110
+ "type": "VISUAL",
111
+ "shape": [
112
+ 3,
113
+ 480,
114
+ 848
115
+ ]
116
+ }
117
+ },
118
+ "output_features": {
119
+ "action": {
120
+ "type": "ACTION",
121
+ "shape": [
122
+ 27
123
+ ]
124
+ }
125
+ },
126
+ "device": "cuda",
127
+ "use_amp": false,
128
+ "use_peft": false,
129
+ "push_to_hub": true,
130
+ "repo_id": "ewayWu/diffusion_policy_pickup_takeout",
131
+ "private": null,
132
+ "tags": null,
133
+ "license": null,
134
+ "pretrained_path": null,
135
+ "horizon": 16,
136
+ "n_action_steps": 8,
137
+ "normalization_mapping": {
138
+ "VISUAL": "MEAN_STD",
139
+ "STATE": "MIN_MAX",
140
+ "ACTION": "MIN_MAX"
141
+ },
142
+ "drop_n_last_frames": 7,
143
+ "vision_backbone": "resnet18",
144
+ "crop_shape": [
145
+ 84,
146
+ 84
147
+ ],
148
+ "crop_is_random": true,
149
+ "pretrained_backbone_weights": null,
150
+ "use_group_norm": true,
151
+ "spatial_softmax_num_keypoints": 32,
152
+ "use_separate_rgb_encoder_per_camera": false,
153
+ "down_dims": [
154
+ 512,
155
+ 1024,
156
+ 2048
157
+ ],
158
+ "kernel_size": 5,
159
+ "n_groups": 8,
160
+ "diffusion_step_embed_dim": 128,
161
+ "use_film_scale_modulation": true,
162
+ "noise_scheduler_type": "DDPM",
163
+ "num_train_timesteps": 100,
164
+ "beta_schedule": "squaredcos_cap_v2",
165
+ "beta_start": 0.0001,
166
+ "beta_end": 0.02,
167
+ "prediction_type": "epsilon",
168
+ "clip_sample": true,
169
+ "clip_sample_range": 1.0,
170
+ "num_inference_steps": null,
171
+ "do_mask_loss_for_padding": false,
172
+ "optimizer_lr": 0.0001,
173
+ "optimizer_betas": [
174
+ 0.95,
175
+ 0.999
176
+ ],
177
+ "optimizer_eps": 1e-08,
178
+ "optimizer_weight_decay": 1e-06,
179
+ "scheduler_name": "cosine",
180
+ "scheduler_warmup_steps": 500
181
+ },
182
+ "output_dir": "outputs/train/diffusion_woosh_pickup_takeout",
183
+ "job_name": "diffusion_woosh_pickup_takeout",
184
+ "resume": false,
185
+ "seed": 1000,
186
+ "num_workers": 4,
187
+ "batch_size": 8,
188
+ "steps": 100000,
189
+ "eval_freq": 20000,
190
+ "log_freq": 200,
191
+ "tolerance_s": 0.0001,
192
+ "save_checkpoint": true,
193
+ "save_freq": 20000,
194
+ "use_policy_training_preset": true,
195
+ "optimizer": {
196
+ "type": "adam",
197
+ "lr": 0.0001,
198
+ "weight_decay": 1e-06,
199
+ "grad_clip_norm": 10.0,
200
+ "betas": [
201
+ 0.95,
202
+ 0.999
203
+ ],
204
+ "eps": 1e-08
205
+ },
206
+ "scheduler": {
207
+ "type": "diffuser",
208
+ "num_warmup_steps": 500,
209
+ "name": "cosine"
210
+ },
211
+ "eval": {
212
+ "n_episodes": 50,
213
+ "batch_size": 50,
214
+ "use_async_envs": false
215
+ },
216
+ "wandb": {
217
+ "enable": true,
218
+ "disable_artifact": false,
219
+ "project": "lerobot",
220
+ "entity": null,
221
+ "notes": null,
222
+ "run_id": "f1attzih",
223
+ "mode": null
224
+ },
225
+ "peft": null,
226
+ "use_rabc": false,
227
+ "rabc_progress_path": null,
228
+ "rabc_kappa": 0.01,
229
+ "rabc_epsilon": 1e-06,
230
+ "rabc_head_mode": "sparse",
231
+ "rename_map": {},
232
+ "checkpoint_path": null
233
+ }