Robotics
LeRobot
Safetensors
act
Muhammad241198 commited on
Commit
59bd66e
·
verified ·
1 Parent(s): b8d1a97

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +69 -0
  3. model.safetensors +3 -0
  4. train_config.json +242 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: lerobot/metaworld_mt50_push_v2_image
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: act
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - robotics
9
+ - act
10
+ - lerobot
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 +1,69 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "act",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 4
9
+ ]
10
+ },
11
+ "observation.environment_state": {
12
+ "type": "ENV",
13
+ "shape": [
14
+ 39
15
+ ]
16
+ },
17
+ "observation.image": {
18
+ "type": "VISUAL",
19
+ "shape": [
20
+ 3,
21
+ 480,
22
+ 480
23
+ ]
24
+ }
25
+ },
26
+ "output_features": {
27
+ "action": {
28
+ "type": "ACTION",
29
+ "shape": [
30
+ 4
31
+ ]
32
+ }
33
+ },
34
+ "device": "cuda",
35
+ "use_amp": false,
36
+ "use_peft": false,
37
+ "push_to_hub": true,
38
+ "repo_id": "Muhammad241198/metaworld-push",
39
+ "private": null,
40
+ "tags": null,
41
+ "license": null,
42
+ "pretrained_path": null,
43
+ "chunk_size": 100,
44
+ "n_action_steps": 100,
45
+ "normalization_mapping": {
46
+ "VISUAL": "MEAN_STD",
47
+ "STATE": "MEAN_STD",
48
+ "ACTION": "MEAN_STD"
49
+ },
50
+ "vision_backbone": "resnet18",
51
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
52
+ "replace_final_stride_with_dilation": false,
53
+ "pre_norm": false,
54
+ "dim_model": 512,
55
+ "n_heads": 8,
56
+ "dim_feedforward": 3200,
57
+ "feedforward_activation": "relu",
58
+ "n_encoder_layers": 4,
59
+ "n_decoder_layers": 1,
60
+ "use_vae": true,
61
+ "latent_dim": 32,
62
+ "n_vae_encoder_layers": 4,
63
+ "temporal_ensemble_coeff": null,
64
+ "dropout": 0.1,
65
+ "kl_weight": 10.0,
66
+ "optimizer_lr": 1e-05,
67
+ "optimizer_weight_decay": 0.0001,
68
+ "optimizer_lr_backbone": 1e-05
69
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f34d365477593c363268233751a7c6b4747229900c3a9392b705b64f2e42a14a
3
+ size 206767520
train_config.json ADDED
@@ -0,0 +1,242 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "lerobot/metaworld_mt50_push_v2_image",
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": {
84
+ "type": "metaworld",
85
+ "task": "assembly-v3",
86
+ "fps": 80,
87
+ "features": {
88
+ "action": {
89
+ "type": "ACTION",
90
+ "shape": [
91
+ 4
92
+ ]
93
+ },
94
+ "agent_pos": {
95
+ "type": "STATE",
96
+ "shape": [
97
+ 4
98
+ ]
99
+ },
100
+ "pixels/top": {
101
+ "type": "VISUAL",
102
+ "shape": [
103
+ 480,
104
+ 480,
105
+ 3
106
+ ]
107
+ }
108
+ },
109
+ "features_map": {
110
+ "action": "action",
111
+ "agent_pos": "observation.state",
112
+ "top": "observation.image",
113
+ "pixels/top": "observation.image"
114
+ },
115
+ "max_parallel_tasks": 1,
116
+ "disable_env_checker": true,
117
+ "episode_length": 400,
118
+ "obs_type": "pixels_agent_pos",
119
+ "render_mode": "rgb_array",
120
+ "multitask_eval": true
121
+ },
122
+ "policy": {
123
+ "type": "act",
124
+ "n_obs_steps": 1,
125
+ "input_features": {
126
+ "observation.state": {
127
+ "type": "STATE",
128
+ "shape": [
129
+ 4
130
+ ]
131
+ },
132
+ "observation.environment_state": {
133
+ "type": "ENV",
134
+ "shape": [
135
+ 39
136
+ ]
137
+ },
138
+ "observation.image": {
139
+ "type": "VISUAL",
140
+ "shape": [
141
+ 3,
142
+ 480,
143
+ 480
144
+ ]
145
+ }
146
+ },
147
+ "output_features": {
148
+ "action": {
149
+ "type": "ACTION",
150
+ "shape": [
151
+ 4
152
+ ]
153
+ }
154
+ },
155
+ "device": "cuda",
156
+ "use_amp": false,
157
+ "use_peft": false,
158
+ "push_to_hub": true,
159
+ "repo_id": "Muhammad241198/metaworld-push",
160
+ "private": null,
161
+ "tags": null,
162
+ "license": null,
163
+ "pretrained_path": null,
164
+ "chunk_size": 100,
165
+ "n_action_steps": 100,
166
+ "normalization_mapping": {
167
+ "VISUAL": "MEAN_STD",
168
+ "STATE": "MEAN_STD",
169
+ "ACTION": "MEAN_STD"
170
+ },
171
+ "vision_backbone": "resnet18",
172
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
173
+ "replace_final_stride_with_dilation": false,
174
+ "pre_norm": false,
175
+ "dim_model": 512,
176
+ "n_heads": 8,
177
+ "dim_feedforward": 3200,
178
+ "feedforward_activation": "relu",
179
+ "n_encoder_layers": 4,
180
+ "n_decoder_layers": 1,
181
+ "use_vae": true,
182
+ "latent_dim": 32,
183
+ "n_vae_encoder_layers": 4,
184
+ "temporal_ensemble_coeff": null,
185
+ "dropout": 0.1,
186
+ "kl_weight": 10.0,
187
+ "optimizer_lr": 1e-05,
188
+ "optimizer_weight_decay": 0.0001,
189
+ "optimizer_lr_backbone": 1e-05
190
+ },
191
+ "output_dir": "outputs",
192
+ "job_name": "metaworld_act",
193
+ "resume": false,
194
+ "seed": 1000,
195
+ "cudnn_deterministic": false,
196
+ "num_workers": 4,
197
+ "batch_size": 4,
198
+ "prefetch_factor": 4,
199
+ "persistent_workers": true,
200
+ "steps": 10,
201
+ "eval_freq": 1000,
202
+ "log_freq": 200,
203
+ "tolerance_s": 0.0001,
204
+ "save_checkpoint": true,
205
+ "save_freq": 20000,
206
+ "use_policy_training_preset": true,
207
+ "optimizer": {
208
+ "type": "adamw",
209
+ "lr": 1e-05,
210
+ "weight_decay": 0.0001,
211
+ "grad_clip_norm": 10.0,
212
+ "betas": [
213
+ 0.9,
214
+ 0.999
215
+ ],
216
+ "eps": 1e-08
217
+ },
218
+ "scheduler": null,
219
+ "eval": {
220
+ "n_episodes": 1,
221
+ "batch_size": 1,
222
+ "use_async_envs": true
223
+ },
224
+ "wandb": {
225
+ "enable": false,
226
+ "disable_artifact": false,
227
+ "project": "lerobot",
228
+ "entity": null,
229
+ "notes": null,
230
+ "run_id": null,
231
+ "mode": null,
232
+ "add_tags": true
233
+ },
234
+ "peft": null,
235
+ "use_rabc": false,
236
+ "rabc_progress_path": null,
237
+ "rabc_kappa": 0.01,
238
+ "rabc_epsilon": 1e-06,
239
+ "rabc_head_mode": "sparse",
240
+ "rename_map": {},
241
+ "checkpoint_path": null
242
+ }