Hugging Face's logo

Mimic-Robotics
/
act_odin_red_x_30a_100k

Robotics
LeRobot
Safetensors
act
Model card Files
xet
Community
ac-pate commited on
Commit
4a2aa7c
·
verified ·
1 Parent(s): 05b1eea

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +97 -0
  2. config.json +87 -0
  3. model.safetensors +3 -0
  4. train_config.json +253 -0
README.md ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets:
3
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v2
4
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v3
5
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v4
6
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v5
7
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v6
8
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v8
9
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v9
10
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v11
11
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v12
12
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v13
13
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v14
14
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v15
15
+ - Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v16
16
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_topleft_v1
17
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_topleft_v2
18
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_toprght_v1
19
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_toprght_v2
20
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_toprght_v3
21
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_toprght_v4
22
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_bottomleft_v1
23
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_bottomleft_v2
24
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_bottomright_v1
25
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_bottomright_v2
26
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_topmiddle_v1
27
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_topMiddle_v3
28
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_topMiddle_v4
29
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_middleLeft_v1
30
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_middleLeft_v2
31
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_middleLeft_v3
32
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_middleLeft_v4
33
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_middleRight_v1
34
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_middleRight_v2
35
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_bottom_middle_v1
36
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_bottom_middle_v2
37
+ - Mimic-Robotics/mimic_tictactoe_red_x_handover_center_slow_v1
38
+ library_name: lerobot
39
+ license: apache-2.0
40
+ model_name: act
41
+ pipeline_tag: robotics
42
+ tags:
43
+ - robotics
44
+ - lerobot
45
+ - act
46
+ ---
47
+
48
+ # Model Card for act
49
+
50
+ <!-- Provide a quick summary of what the model is/does. -->
51
+
52
+
53
+ [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.
54
+
55
+
56
+ This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
57
+ See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).
58
+
59
+ ---
60
+
61
+ ## How to Get Started with the Model
62
+
63
+ For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
64
+ Below is the short version on how to train and run inference/eval:
65
+
66
+ ### Train from scratch
67
+
68
+ ```bash
69
+ lerobot-train \
70
+ --dataset.repo_id=${HF_USER}/<dataset> \
71
+ --policy.type=act \
72
+ --output_dir=outputs/train/<desired_policy_repo_id> \
73
+ --job_name=lerobot_training \
74
+ --policy.device=cuda \
75
+ --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
76
+ --wandb.enable=true
77
+ ```
78
+
79
+ _Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`._
80
+
81
+ ### Evaluate the policy/run inference
82
+
83
+ ```bash
84
+ lerobot-record \
85
+ --robot.type=so100_follower \
86
+ --dataset.repo_id=<hf_user>/eval_<dataset> \
87
+ --policy.path=<hf_user>/<desired_policy_repo_id> \
88
+ --episodes=10
89
+ ```
90
+
91
+ Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.
92
+
93
+ ---
94
+
95
+ ## Model Details
96
+
97
+ - **License:** apache-2.0
config.json ADDED
@@ -0,0 +1,87 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "act",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 15
9
+ ]
10
+ },
11
+ "observation.images.right_wrist": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 480,
16
+ 640
17
+ ]
18
+ },
19
+ "observation.images.left_wrist": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 480,
24
+ 640
25
+ ]
26
+ },
27
+ "observation.images.front": {
28
+ "type": "VISUAL",
29
+ "shape": [
30
+ 3,
31
+ 480,
32
+ 640
33
+ ]
34
+ },
35
+ "observation.images.top": {
36
+ "type": "VISUAL",
37
+ "shape": [
38
+ 3,
39
+ 720,
40
+ 1280
41
+ ]
42
+ }
43
+ },
44
+ "output_features": {
45
+ "action": {
46
+ "type": "ACTION",
47
+ "shape": [
48
+ 15
49
+ ]
50
+ }
51
+ },
52
+ "device": "cuda",
53
+ "use_amp": false,
54
+ "use_peft": false,
55
+ "push_to_hub": true,
56
+ "repo_id": "Mimic-Robotics/act_odin_red_x_handover_and_place_tictactoe_v8",
57
+ "private": null,
58
+ "tags": null,
59
+ "license": null,
60
+ "pretrained_path": null,
61
+ "chunk_size": 100,
62
+ "n_action_steps": 30,
63
+ "normalization_mapping": {
64
+ "VISUAL": "MEAN_STD",
65
+ "STATE": "MEAN_STD",
66
+ "ACTION": "MEAN_STD"
67
+ },
68
+ "vision_backbone": "resnet18",
69
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
70
+ "replace_final_stride_with_dilation": false,
71
+ "pre_norm": false,
72
+ "dim_model": 512,
73
+ "n_heads": 8,
74
+ "dim_feedforward": 3200,
75
+ "feedforward_activation": "relu",
76
+ "n_encoder_layers": 4,
77
+ "n_decoder_layers": 1,
78
+ "use_vae": true,
79
+ "latent_dim": 32,
80
+ "n_vae_encoder_layers": 4,
81
+ "temporal_ensemble_coeff": null,
82
+ "dropout": 0.1,
83
+ "kl_weight": 10.0,
84
+ "optimizer_lr": 1e-05,
85
+ "optimizer_weight_decay": 0.0001,
86
+ "optimizer_lr_backbone": 1e-05
87
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f63bf449229d7a8102e583a3624758880ce7af8f2f62b50877f8850a8bd43341
3
+ size 206773500
train_config.json ADDED
@@ -0,0 +1,253 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": [
4
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v2",
5
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v3",
6
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v4",
7
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v5",
8
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v6",
9
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v8",
10
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v9",
11
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v11",
12
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v12",
13
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v13",
14
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v14",
15
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v15",
16
+ "Mimic-Robotics/mimic_tictactoe_pick_red_x_handover_place_mm_v16",
17
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_topleft_v1",
18
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_topleft_v2",
19
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_toprght_v1",
20
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_toprght_v2",
21
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_toprght_v3",
22
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_toprght_v4",
23
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_bottomleft_v1",
24
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_bottomleft_v2",
25
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_bottomright_v1",
26
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_bottomright_v2",
27
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_topmiddle_v1",
28
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_topMiddle_v3",
29
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_topMiddle_v4",
30
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_middleLeft_v1",
31
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_middleLeft_v2",
32
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_middleLeft_v3",
33
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_middleLeft_v4",
34
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_middleRight_v1",
35
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_middleRight_v2",
36
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_bottom_middle_v1",
37
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_bottom_middle_v2",
38
+ "Mimic-Robotics/mimic_tictactoe_red_x_handover_center_slow_v1"
39
+ ],
40
+ "root": null,
41
+ "episodes": null,
42
+ "image_transforms": {
43
+ "enable": false,
44
+ "max_num_transforms": 3,
45
+ "random_order": false,
46
+ "tfs": {
47
+ "brightness": {
48
+ "weight": 1.0,
49
+ "type": "ColorJitter",
50
+ "kwargs": {
51
+ "brightness": [
52
+ 0.8,
53
+ 1.2
54
+ ]
55
+ }
56
+ },
57
+ "contrast": {
58
+ "weight": 1.0,
59
+ "type": "ColorJitter",
60
+ "kwargs": {
61
+ "contrast": [
62
+ 0.8,
63
+ 1.2
64
+ ]
65
+ }
66
+ },
67
+ "saturation": {
68
+ "weight": 1.0,
69
+ "type": "ColorJitter",
70
+ "kwargs": {
71
+ "saturation": [
72
+ 0.5,
73
+ 1.5
74
+ ]
75
+ }
76
+ },
77
+ "hue": {
78
+ "weight": 1.0,
79
+ "type": "ColorJitter",
80
+ "kwargs": {
81
+ "hue": [
82
+ -0.05,
83
+ 0.05
84
+ ]
85
+ }
86
+ },
87
+ "sharpness": {
88
+ "weight": 1.0,
89
+ "type": "SharpnessJitter",
90
+ "kwargs": {
91
+ "sharpness": [
92
+ 0.5,
93
+ 1.5
94
+ ]
95
+ }
96
+ },
97
+ "affine": {
98
+ "weight": 1.0,
99
+ "type": "RandomAffine",
100
+ "kwargs": {
101
+ "degrees": [
102
+ -5.0,
103
+ 5.0
104
+ ],
105
+ "translate": [
106
+ 0.05,
107
+ 0.05
108
+ ]
109
+ }
110
+ }
111
+ }
112
+ },
113
+ "revision": null,
114
+ "use_imagenet_stats": true,
115
+ "video_backend": "torchcodec",
116
+ "streaming": false
117
+ },
118
+ "env": null,
119
+ "policy": {
120
+ "type": "act",
121
+ "n_obs_steps": 1,
122
+ "input_features": {
123
+ "observation.state": {
124
+ "type": "STATE",
125
+ "shape": [
126
+ 15
127
+ ]
128
+ },
129
+ "observation.images.right_wrist": {
130
+ "type": "VISUAL",
131
+ "shape": [
132
+ 3,
133
+ 480,
134
+ 640
135
+ ]
136
+ },
137
+ "observation.images.left_wrist": {
138
+ "type": "VISUAL",
139
+ "shape": [
140
+ 3,
141
+ 480,
142
+ 640
143
+ ]
144
+ },
145
+ "observation.images.front": {
146
+ "type": "VISUAL",
147
+ "shape": [
148
+ 3,
149
+ 480,
150
+ 640
151
+ ]
152
+ },
153
+ "observation.images.top": {
154
+ "type": "VISUAL",
155
+ "shape": [
156
+ 3,
157
+ 720,
158
+ 1280
159
+ ]
160
+ }
161
+ },
162
+ "output_features": {
163
+ "action": {
164
+ "type": "ACTION",
165
+ "shape": [
166
+ 15
167
+ ]
168
+ }
169
+ },
170
+ "device": "cuda",
171
+ "use_amp": false,
172
+ "use_peft": false,
173
+ "push_to_hub": true,
174
+ "repo_id": "Mimic-Robotics/act_odin_red_x_handover_and_place_tictactoe_v8",
175
+ "private": null,
176
+ "tags": null,
177
+ "license": null,
178
+ "pretrained_path": null,
179
+ "chunk_size": 100,
180
+ "n_action_steps": 30,
181
+ "normalization_mapping": {
182
+ "VISUAL": "MEAN_STD",
183
+ "STATE": "MEAN_STD",
184
+ "ACTION": "MEAN_STD"
185
+ },
186
+ "vision_backbone": "resnet18",
187
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
188
+ "replace_final_stride_with_dilation": false,
189
+ "pre_norm": false,
190
+ "dim_model": 512,
191
+ "n_heads": 8,
192
+ "dim_feedforward": 3200,
193
+ "feedforward_activation": "relu",
194
+ "n_encoder_layers": 4,
195
+ "n_decoder_layers": 1,
196
+ "use_vae": true,
197
+ "latent_dim": 32,
198
+ "n_vae_encoder_layers": 4,
199
+ "temporal_ensemble_coeff": null,
200
+ "dropout": 0.1,
201
+ "kl_weight": 10.0,
202
+ "optimizer_lr": 1e-05,
203
+ "optimizer_weight_decay": 0.0001,
204
+ "optimizer_lr_backbone": 1e-05
205
+ },
206
+ "output_dir": "/home/odin/mimic-lerobot/outputs/train/act_odin_red_x_handover_and_place_tictactoe_v8",
207
+ "job_name": "act_odin_red_x_handover_and_place_tictactoe_v8",
208
+ "resume": false,
209
+ "seed": 1000,
210
+ "num_workers": 12,
211
+ "batch_size": 10,
212
+ "steps": 100000,
213
+ "eval_freq": 20000,
214
+ "log_freq": 200,
215
+ "tolerance_s": 0.0001,
216
+ "save_checkpoint": true,
217
+ "save_freq": 10000,
218
+ "use_policy_training_preset": true,
219
+ "optimizer": {
220
+ "type": "adamw",
221
+ "lr": 1e-05,
222
+ "weight_decay": 0.0001,
223
+ "grad_clip_norm": 10.0,
224
+ "betas": [
225
+ 0.9,
226
+ 0.999
227
+ ],
228
+ "eps": 1e-08
229
+ },
230
+ "scheduler": null,
231
+ "eval": {
232
+ "n_episodes": 50,
233
+ "batch_size": 50,
234
+ "use_async_envs": false
235
+ },
236
+ "wandb": {
237
+ "enable": true,
238
+ "disable_artifact": false,
239
+ "project": "lerobot",
240
+ "entity": null,
241
+ "notes": "Multi-dataset ACT training on red_x_handover_and_place_tictactoe_v8 with Computer odin and Batch 10",
242
+ "run_id": "l5ybc6d8",
243
+ "mode": null
244
+ },
245
+ "peft": null,
246
+ "use_rabc": false,
247
+ "rabc_progress_path": null,
248
+ "rabc_kappa": 0.01,
249
+ "rabc_epsilon": 1e-06,
250
+ "rabc_head_mode": "sparse",
251
+ "rename_map": {},
252
+ "checkpoint_path": null
253
+ }