Hugging Face's logo

iantc104
/
sim_hook_package_3arms_zed_act

Robotics
Safetensors
lerobot
act
model_hub_mixin
pytorch_model_hub_mixin
Model card Files Community
iantc104 commited on
Commit
ec55e48
•
1 Parent(s): e4420ca

Upload YAML configuration

Browse files
Files changed (1) hide show
  1. config.yaml +185 -0
config.yaml ADDED
@@ -0,0 +1,185 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ resume: false
2
+ device: cuda
3
+ use_amp: false
4
+ seed: 1000
5
+ dataset_repo_id: iantc104/gv_sim_hook_package_3arms
6
+ video_backend: pyav
7
+ training:
8
+ offline_steps: 15625
9
+ num_workers: 4
10
+ batch_size: 16
11
+ eval_freq: 1000
12
+ log_freq: 100
13
+ save_checkpoint: true
14
+ save_freq: 1000
15
+ online_steps: 0
16
+ online_rollout_n_episodes: 1
17
+ online_rollout_batch_size: 1
18
+ online_steps_between_rollouts: 1
19
+ online_sampling_ratio: 0.5
20
+ online_env_seed: null
21
+ online_buffer_capacity: null
22
+ online_buffer_seed_size: 0
23
+ do_online_rollout_async: false
24
+ image_transforms:
25
+ enable: false
26
+ max_num_transforms: 3
27
+ random_order: false
28
+ brightness:
29
+ weight: 1
30
+ min_max:
31
+ - 0.8
32
+ - 1.2
33
+ contrast:
34
+ weight: 1
35
+ min_max:
36
+ - 0.8
37
+ - 1.2
38
+ saturation:
39
+ weight: 1
40
+ min_max:
41
+ - 0.5
42
+ - 1.5
43
+ hue:
44
+ weight: 1
45
+ min_max:
46
+ - -0.05
47
+ - 0.05
48
+ sharpness:
49
+ weight: 1
50
+ min_max:
51
+ - 0.8
52
+ - 1.2
53
+ lr: 2.5e-05
54
+ lr_backbone: 1.0e-05
55
+ weight_decay: 0.0001
56
+ grad_clip_norm: 10
57
+ delta_timestamps:
58
+ action:
59
+ - 0.0
60
+ - 0.04
61
+ - 0.08
62
+ - 0.12
63
+ - 0.16
64
+ - 0.2
65
+ - 0.24
66
+ - 0.28
67
+ - 0.32
68
+ - 0.36
69
+ - 0.4
70
+ - 0.44
71
+ - 0.48
72
+ - 0.52
73
+ - 0.56
74
+ - 0.6
75
+ - 0.64
76
+ - 0.68
77
+ - 0.72
78
+ - 0.76
79
+ - 0.8
80
+ - 0.84
81
+ - 0.88
82
+ - 0.92
83
+ - 0.96
84
+ - 1.0
85
+ - 1.04
86
+ - 1.08
87
+ - 1.12
88
+ - 1.16
89
+ - 1.2
90
+ - 1.24
91
+ - 1.28
92
+ - 1.32
93
+ - 1.36
94
+ - 1.4
95
+ - 1.44
96
+ - 1.48
97
+ - 1.52
98
+ - 1.56
99
+ - 1.6
100
+ - 1.64
101
+ - 1.68
102
+ - 1.72
103
+ - 1.76
104
+ - 1.8
105
+ - 1.84
106
+ - 1.88
107
+ - 1.92
108
+ - 1.96
109
+ eval:
110
+ n_episodes: 1
111
+ batch_size: 1
112
+ use_async_envs: false
113
+ wandb:
114
+ enable: true
115
+ disable_artifact: false
116
+ project: lerobot
117
+ notes: ''
118
+ fps: 25
119
+ env:
120
+ name: guided_vision
121
+ task: HookPackage-3Arms-v0
122
+ state_dim: 21
123
+ action_dim: 21
124
+ fps: ${fps}
125
+ episode_length: 300
126
+ override_dataset_stats:
127
+ observation.images.zed_cam_left:
128
+ mean:
129
+ - - - 0.485
130
+ - - - 0.456
131
+ - - - 0.406
132
+ std:
133
+ - - - 0.229
134
+ - - - 0.224
135
+ - - - 0.225
136
+ observation.images.zed_cam_right:
137
+ mean:
138
+ - - - 0.485
139
+ - - - 0.456
140
+ - - - 0.406
141
+ std:
142
+ - - - 0.229
143
+ - - - 0.224
144
+ - - - 0.225
145
+ policy:
146
+ name: act
147
+ n_obs_steps: 1
148
+ chunk_size: 50
149
+ n_action_steps: 50
150
+ input_shapes:
151
+ observation.images.zed_cam_left:
152
+ - 3
153
+ - 480
154
+ - 640
155
+ observation.images.zed_cam_right:
156
+ - 3
157
+ - 480
158
+ - 640
159
+ observation.state:
160
+ - ${env.state_dim}
161
+ output_shapes:
162
+ action:
163
+ - ${env.action_dim}
164
+ input_normalization_modes:
165
+ observation.images.zed_cam_left: mean_std
166
+ observation.images.zed_cam_right: mean_std
167
+ observation.state: mean_std
168
+ output_normalization_modes:
169
+ action: mean_std
170
+ vision_backbone: resnet18
171
+ pretrained_backbone_weights: ResNet18_Weights.IMAGENET1K_V1
172
+ replace_final_stride_with_dilation: false
173
+ pre_norm: false
174
+ dim_model: 512
175
+ n_heads: 8
176
+ dim_feedforward: 3200
177
+ feedforward_activation: relu
178
+ n_encoder_layers: 4
179
+ n_decoder_layers: 1
180
+ use_vae: true
181
+ latent_dim: 32
182
+ n_vae_encoder_layers: 4
183
+ temporal_ensemble_coeff: null
184
+ dropout: 0.1
185
+ kl_weight: 10.0