Hugging Face's logo

andabi
/
M16

Robotics
LeRobot
Safetensors
act
model_hub_mixin
pytorch_model_hub_mixin
Model card Files
xet
 Community
andabi commited on
Commit
798bf73
·
verified ·
1 Parent(s): c1b82d6

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. README.md +12 -0
  2. config.json +55 -0
  3. config.yaml +183 -0
  4. model.safetensors +3 -0
README.md ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: lerobot
3
+ tags:
4
+ - act
5
+ - model_hub_mixin
6
+ - pytorch_model_hub_mixin
7
+ - robotics
8
+ ---
9
+
10
+ This model has been pushed to the Hub using the [PytorchModelHubMixin](https://huggingface.co/docs/huggingface_hub/package_reference/mixins#huggingface_hub.PyTorchModelHubMixin) integration:
11
+ - Library: https://github.com/huggingface/lerobot
12
+ - Docs: [More Information Needed]
config.json ADDED
@@ -0,0 +1,55 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "chunk_size": 30,
3
+ "dim_feedforward": 3200,
4
+ "dim_model": 512,
5
+ "dropout": 0.1,
6
+ "feedforward_activation": "relu",
7
+ "input_normalization_modes": {
8
+ "observation.images.bird": "mean_std",
9
+ "observation.images.wrist_left": "mean_std",
10
+ "observation.images.wrist_right": "mean_std",
11
+ "observation.state": "mean_std"
12
+ },
13
+ "input_shapes": {
14
+ "observation.images.bird": [
15
+ 3,
16
+ 480,
17
+ 640
18
+ ],
19
+ "observation.images.wrist_left": [
20
+ 3,
21
+ 480,
22
+ 640
23
+ ],
24
+ "observation.images.wrist_right": [
25
+ 3,
26
+ 480,
27
+ 640
28
+ ],
29
+ "observation.state": [
30
+ 12
31
+ ]
32
+ },
33
+ "kl_weight": 10.0,
34
+ "latent_dim": 32,
35
+ "n_action_steps": 30,
36
+ "n_decoder_layers": 1,
37
+ "n_encoder_layers": 4,
38
+ "n_heads": 8,
39
+ "n_obs_steps": 1,
40
+ "n_vae_encoder_layers": 4,
41
+ "output_normalization_modes": {
42
+ "action": "mean_std"
43
+ },
44
+ "output_shapes": {
45
+ "action": [
46
+ 12
47
+ ]
48
+ },
49
+ "pre_norm": false,
50
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
51
+ "replace_final_stride_with_dilation": false,
52
+ "temporal_ensemble_coeff": null,
53
+ "use_vae": true,
54
+ "vision_backbone": "resnet18"
55
+ }
config.yaml ADDED
@@ -0,0 +1,183 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ resume: true
2
+ device: cuda
3
+ use_amp: false
4
+ seed: 1000
5
+ dataset_repo_id:
6
+ - andabi/D8
7
+ - andabi/D12
8
+ - andabi/D15
9
+ - andabi/D16
10
+ - andabi/D17
11
+ video_backend: pyav
12
+ training:
13
+ offline_steps: 2000000
14
+ num_workers: 4
15
+ batch_size: 16
16
+ eval_freq: -1
17
+ log_freq: 100
18
+ save_checkpoint: true
19
+ save_freq: 10000
20
+ online_steps: 0
21
+ online_rollout_n_episodes: 1
22
+ online_rollout_batch_size: 1
23
+ online_steps_between_rollouts: 1
24
+ online_sampling_ratio: 0.5
25
+ online_env_seed: null
26
+ online_buffer_capacity: null
27
+ online_buffer_seed_size: 0
28
+ do_online_rollout_async: false
29
+ image_transforms:
30
+ enable: false
31
+ max_num_transforms: 3
32
+ random_order: false
33
+ brightness:
34
+ weight: 1
35
+ min_max:
36
+ - 0.8
37
+ - 1.2
38
+ contrast:
39
+ weight: 1
40
+ min_max:
41
+ - 0.8
42
+ - 1.2
43
+ saturation:
44
+ weight: 1
45
+ min_max:
46
+ - 0.5
47
+ - 1.5
48
+ hue:
49
+ weight: 1
50
+ min_max:
51
+ - -0.05
52
+ - 0.05
53
+ sharpness:
54
+ weight: 1
55
+ min_max:
56
+ - 0.8
57
+ - 1.2
58
+ lr: 1.0e-05
59
+ lr_backbone: 1.0e-05
60
+ weight_decay: 0.0001
61
+ grad_clip_norm: 10
62
+ delta_timestamps:
63
+ action:
64
+ - 0.0
65
+ - 0.06666666666666667
66
+ - 0.13333333333333333
67
+ - 0.2
68
+ - 0.26666666666666666
69
+ - 0.3333333333333333
70
+ - 0.4
71
+ - 0.4666666666666667
72
+ - 0.5333333333333333
73
+ - 0.6
74
+ - 0.6666666666666666
75
+ - 0.7333333333333333
76
+ - 0.8
77
+ - 0.8666666666666667
78
+ - 0.9333333333333333
79
+ - 1.0
80
+ - 1.0666666666666667
81
+ - 1.1333333333333333
82
+ - 1.2
83
+ - 1.2666666666666666
84
+ - 1.3333333333333333
85
+ - 1.4
86
+ - 1.4666666666666666
87
+ - 1.5333333333333334
88
+ - 1.6
89
+ - 1.6666666666666667
90
+ - 1.7333333333333334
91
+ - 1.8
92
+ - 1.8666666666666667
93
+ - 1.9333333333333333
94
+ eval:
95
+ n_episodes: 50
96
+ batch_size: 50
97
+ use_async_envs: false
98
+ wandb:
99
+ enable: true
100
+ disable_artifact: false
101
+ project: lerobot
102
+ notes: ''
103
+ fps: 15
104
+ env:
105
+ name: real_world
106
+ task: null
107
+ state_dim: 12
108
+ action_dim: 12
109
+ fps: ${fps}
110
+ override_dataset_stats:
111
+ observation.images.bird:
112
+ mean:
113
+ - - - 0.485
114
+ - - - 0.456
115
+ - - - 0.406
116
+ std:
117
+ - - - 0.229
118
+ - - - 0.224
119
+ - - - 0.225
120
+ observation.images.wrist_left:
121
+ mean:
122
+ - - - 0.485
123
+ - - - 0.456
124
+ - - - 0.406
125
+ std:
126
+ - - - 0.229
127
+ - - - 0.224
128
+ - - - 0.225
129
+ observation.images.wrist_right:
130
+ mean:
131
+ - - - 0.485
132
+ - - - 0.456
133
+ - - - 0.406
134
+ std:
135
+ - - - 0.229
136
+ - - - 0.224
137
+ - - - 0.225
138
+ policy:
139
+ name: act
140
+ n_obs_steps: 1
141
+ chunk_size: 30
142
+ n_action_steps: 30
143
+ input_shapes:
144
+ observation.images.bird:
145
+ - 3
146
+ - 480
147
+ - 640
148
+ observation.images.wrist_left:
149
+ - 3
150
+ - 480
151
+ - 640
152
+ observation.images.wrist_right:
153
+ - 3
154
+ - 480
155
+ - 640
156
+ observation.state:
157
+ - ${env.state_dim}
158
+ output_shapes:
159
+ action:
160
+ - ${env.action_dim}
161
+ input_normalization_modes:
162
+ observation.images.bird: mean_std
163
+ observation.images.wrist_left: mean_std
164
+ observation.images.wrist_right: mean_std
165
+ observation.state: mean_std
166
+ output_normalization_modes:
167
+ action: mean_std
168
+ vision_backbone: resnet18
169
+ pretrained_backbone_weights: ResNet18_Weights.IMAGENET1K_V1
170
+ replace_final_stride_with_dilation: false
171
+ pre_norm: false
172
+ dim_model: 512
173
+ n_heads: 8
174
+ dim_feedforward: 3200
175
+ feedforward_activation: relu
176
+ n_encoder_layers: 4
177
+ n_decoder_layers: 1
178
+ use_vae: true
179
+ latent_dim: 32
180
+ n_vae_encoder_layers: 4
181
+ temporal_ensemble_coeff: null
182
+ dropout: 0.1
183
+ kl_weight: 10.0
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6e37c6cca1631948d0aa2627c67eef678adf7f5349760fca63943685b58e81b1
3
+ size 206463968