Hugging Face's logo

seriintan
/
act_spatialPE

Robotics
LeRobot
Safetensors
act
Model card Files
xet
 Community
seriintan commited on
Commit
ce5ef4f
·
verified ·
1 Parent(s): a1f6657

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +136 -0
  3. model.safetensors +3 -0
  4. train_config.json +266 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: seriintan/pick_scissor_spatialPE_v1_enriched
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: act
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - lerobot
9
+ - robotics
10
+ - act
11
+ ---
12
+
13
+ # Model Card for act
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+ [Action Chunking with Transformers (ACT)](https://huggingface.co/papers/2304.13705) is an imitation-learning method that predicts short action chunks instead of single steps. It learns from teleoperated data and often achieves high success rates.
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 +1,136 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "act",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 6
9
+ ]
10
+ },
11
+ "observation.images.front": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 480,
16
+ 640
17
+ ]
18
+ },
19
+ "observation.apriltag.basket.x": {
20
+ "type": "STATE",
21
+ "shape": [
22
+ 1
23
+ ]
24
+ },
25
+ "observation.apriltag.basket.y": {
26
+ "type": "STATE",
27
+ "shape": [
28
+ 1
29
+ ]
30
+ },
31
+ "observation.apriltag.basket.z": {
32
+ "type": "STATE",
33
+ "shape": [
34
+ 1
35
+ ]
36
+ },
37
+ "observation.apriltag.basket.valid": {
38
+ "type": "STATE",
39
+ "shape": [
40
+ 1
41
+ ]
42
+ },
43
+ "observation.apriltag.anchor1.x": {
44
+ "type": "STATE",
45
+ "shape": [
46
+ 1
47
+ ]
48
+ },
49
+ "observation.apriltag.anchor1.y": {
50
+ "type": "STATE",
51
+ "shape": [
52
+ 1
53
+ ]
54
+ },
55
+ "observation.apriltag.anchor1.z": {
56
+ "type": "STATE",
57
+ "shape": [
58
+ 1
59
+ ]
60
+ },
61
+ "observation.apriltag.anchor1.valid": {
62
+ "type": "STATE",
63
+ "shape": [
64
+ 1
65
+ ]
66
+ },
67
+ "observation.apriltag.anchor2.x": {
68
+ "type": "STATE",
69
+ "shape": [
70
+ 1
71
+ ]
72
+ },
73
+ "observation.apriltag.anchor2.y": {
74
+ "type": "STATE",
75
+ "shape": [
76
+ 1
77
+ ]
78
+ },
79
+ "observation.apriltag.anchor2.z": {
80
+ "type": "STATE",
81
+ "shape": [
82
+ 1
83
+ ]
84
+ },
85
+ "observation.apriltag.anchor2.valid": {
86
+ "type": "STATE",
87
+ "shape": [
88
+ 1
89
+ ]
90
+ }
91
+ },
92
+ "output_features": {
93
+ "action": {
94
+ "type": "ACTION",
95
+ "shape": [
96
+ 6
97
+ ]
98
+ }
99
+ },
100
+ "device": "cuda",
101
+ "use_amp": false,
102
+ "use_peft": false,
103
+ "push_to_hub": true,
104
+ "repo_id": "seriintan/act_spatialPE",
105
+ "private": null,
106
+ "tags": null,
107
+ "license": null,
108
+ "pretrained_path": null,
109
+ "chunk_size": 100,
110
+ "n_action_steps": 100,
111
+ "normalization_mapping": {
112
+ "VISUAL": "MEAN_STD",
113
+ "STATE": "MEAN_STD",
114
+ "ACTION": "MEAN_STD"
115
+ },
116
+ "vision_backbone": "resnet18",
117
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
118
+ "replace_final_stride_with_dilation": false,
119
+ "pre_norm": false,
120
+ "dim_model": 512,
121
+ "n_heads": 8,
122
+ "dim_feedforward": 3200,
123
+ "feedforward_activation": "relu",
124
+ "n_encoder_layers": 4,
125
+ "n_decoder_layers": 1,
126
+ "use_vae": true,
127
+ "latent_dim": 32,
128
+ "n_vae_encoder_layers": 4,
129
+ "temporal_ensemble_coeff": null,
130
+ "dropout": 0.1,
131
+ "kl_weight": 10.0,
132
+ "use_apriltag_spatial_pe": true,
133
+ "optimizer_lr": 1e-05,
134
+ "optimizer_weight_decay": 0.0001,
135
+ "optimizer_lr_backbone": 1e-05
136
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d979a31489b221aa7e8b5f2e965cd90db17f4a95326232b02cc13ac885a30ca9
3
+ size 207872224
train_config.json ADDED
@@ -0,0 +1,266 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "seriintan/pick_scissor_spatialPE_v1_enriched",
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": "act",
85
+ "n_obs_steps": 1,
86
+ "input_features": {
87
+ "observation.state": {
88
+ "type": "STATE",
89
+ "shape": [
90
+ 6
91
+ ]
92
+ },
93
+ "observation.images.front": {
94
+ "type": "VISUAL",
95
+ "shape": [
96
+ 3,
97
+ 480,
98
+ 640
99
+ ]
100
+ },
101
+ "observation.apriltag.basket.x": {
102
+ "type": "STATE",
103
+ "shape": [
104
+ 1
105
+ ]
106
+ },
107
+ "observation.apriltag.basket.y": {
108
+ "type": "STATE",
109
+ "shape": [
110
+ 1
111
+ ]
112
+ },
113
+ "observation.apriltag.basket.z": {
114
+ "type": "STATE",
115
+ "shape": [
116
+ 1
117
+ ]
118
+ },
119
+ "observation.apriltag.basket.valid": {
120
+ "type": "STATE",
121
+ "shape": [
122
+ 1
123
+ ]
124
+ },
125
+ "observation.apriltag.anchor1.x": {
126
+ "type": "STATE",
127
+ "shape": [
128
+ 1
129
+ ]
130
+ },
131
+ "observation.apriltag.anchor1.y": {
132
+ "type": "STATE",
133
+ "shape": [
134
+ 1
135
+ ]
136
+ },
137
+ "observation.apriltag.anchor1.z": {
138
+ "type": "STATE",
139
+ "shape": [
140
+ 1
141
+ ]
142
+ },
143
+ "observation.apriltag.anchor1.valid": {
144
+ "type": "STATE",
145
+ "shape": [
146
+ 1
147
+ ]
148
+ },
149
+ "observation.apriltag.anchor2.x": {
150
+ "type": "STATE",
151
+ "shape": [
152
+ 1
153
+ ]
154
+ },
155
+ "observation.apriltag.anchor2.y": {
156
+ "type": "STATE",
157
+ "shape": [
158
+ 1
159
+ ]
160
+ },
161
+ "observation.apriltag.anchor2.z": {
162
+ "type": "STATE",
163
+ "shape": [
164
+ 1
165
+ ]
166
+ },
167
+ "observation.apriltag.anchor2.valid": {
168
+ "type": "STATE",
169
+ "shape": [
170
+ 1
171
+ ]
172
+ }
173
+ },
174
+ "output_features": {
175
+ "action": {
176
+ "type": "ACTION",
177
+ "shape": [
178
+ 6
179
+ ]
180
+ }
181
+ },
182
+ "device": "cuda",
183
+ "use_amp": false,
184
+ "use_peft": false,
185
+ "push_to_hub": true,
186
+ "repo_id": "seriintan/act_spatialPE",
187
+ "private": null,
188
+ "tags": null,
189
+ "license": null,
190
+ "pretrained_path": null,
191
+ "chunk_size": 100,
192
+ "n_action_steps": 100,
193
+ "normalization_mapping": {
194
+ "VISUAL": "MEAN_STD",
195
+ "STATE": "MEAN_STD",
196
+ "ACTION": "MEAN_STD"
197
+ },
198
+ "vision_backbone": "resnet18",
199
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
200
+ "replace_final_stride_with_dilation": false,
201
+ "pre_norm": false,
202
+ "dim_model": 512,
203
+ "n_heads": 8,
204
+ "dim_feedforward": 3200,
205
+ "feedforward_activation": "relu",
206
+ "n_encoder_layers": 4,
207
+ "n_decoder_layers": 1,
208
+ "use_vae": true,
209
+ "latent_dim": 32,
210
+ "n_vae_encoder_layers": 4,
211
+ "temporal_ensemble_coeff": null,
212
+ "dropout": 0.1,
213
+ "kl_weight": 10.0,
214
+ "use_apriltag_spatial_pe": true,
215
+ "optimizer_lr": 1e-05,
216
+ "optimizer_weight_decay": 0.0001,
217
+ "optimizer_lr_backbone": 1e-05
218
+ },
219
+ "output_dir": "outputs/train/act_spatialPE",
220
+ "job_name": "act_spatialPE",
221
+ "resume": false,
222
+ "seed": 1000,
223
+ "num_workers": 4,
224
+ "batch_size": 8,
225
+ "steps": 50000,
226
+ "eval_freq": 20000,
227
+ "log_freq": 100,
228
+ "tolerance_s": 0.0001,
229
+ "save_checkpoint": true,
230
+ "save_freq": 10000,
231
+ "use_policy_training_preset": true,
232
+ "optimizer": {
233
+ "type": "adamw",
234
+ "lr": 1e-05,
235
+ "weight_decay": 0.0001,
236
+ "grad_clip_norm": 10.0,
237
+ "betas": [
238
+ 0.9,
239
+ 0.999
240
+ ],
241
+ "eps": 1e-08
242
+ },
243
+ "scheduler": null,
244
+ "eval": {
245
+ "n_episodes": 50,
246
+ "batch_size": 50,
247
+ "use_async_envs": false
248
+ },
249
+ "wandb": {
250
+ "enable": false,
251
+ "disable_artifact": false,
252
+ "project": "lerobot",
253
+ "entity": null,
254
+ "notes": null,
255
+ "run_id": null,
256
+ "mode": null
257
+ },
258
+ "peft": null,
259
+ "use_rabc": false,
260
+ "rabc_progress_path": null,
261
+ "rabc_kappa": 0.01,
262
+ "rabc_epsilon": 1e-06,
263
+ "rabc_head_mode": "sparse",
264
+ "rename_map": {},
265
+ "checkpoint_path": null
266
+ }