Hugging Face's logo

jhseon-holiday
/
vqbet_xarm_holi_test_175000

Robotics
Safetensors
lerobot
model_hub_mixin
pytorch_model_hub_mixin
vqbet
Model card Files Community
jhseon-holiday commited on
Commit
9dd0310
•
1 Parent(s): 89a171a

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. README.md +12 -0
  2. config.json +65 -0
  3. config.yaml +210 -0
  4. model.safetensors +3 -0
README.md ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: lerobot
3
+ tags:
4
+ - model_hub_mixin
5
+ - pytorch_model_hub_mixin
6
+ - robotics
7
+ - vqbet
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 +1,65 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action_chunk_size": 5,
3
+ "bet_softmax_temperature": 0.1,
4
+ "crop_is_random": true,
5
+ "crop_shape": [
6
+ 84,
7
+ 84
8
+ ],
9
+ "dropout": 0.1,
10
+ "gpt_block_size": 500,
11
+ "gpt_hidden_dim": 512,
12
+ "gpt_input_dim": 512,
13
+ "gpt_n_head": 8,
14
+ "gpt_n_layer": 8,
15
+ "gpt_output_dim": 512,
16
+ "input_normalization_modes": {
17
+ "observation.image.cam_left": "mean_std",
18
+ "observation.image.cam_right": "mean_std",
19
+ "observation.image.cam_wrist": "mean_std",
20
+ "observation.state": "mean_std"
21
+ },
22
+ "input_shapes": {
23
+ "observation.image.cam_left": [
24
+ 3,
25
+ 96,
26
+ 96
27
+ ],
28
+ "observation.image.cam_right": [
29
+ 3,
30
+ 96,
31
+ 96
32
+ ],
33
+ "observation.image.cam_wrist": [
34
+ 3,
35
+ 96,
36
+ 96
37
+ ],
38
+ "observation.state": [
39
+ 7
40
+ ]
41
+ },
42
+ "mlp_hidden_dim": 1024,
43
+ "n_action_pred_token": 7,
44
+ "n_obs_steps": 5,
45
+ "n_vqvae_training_steps": 20000,
46
+ "offset_loss_weight": 10000.0,
47
+ "output_normalization_modes": {
48
+ "action": "mean_std"
49
+ },
50
+ "output_shapes": {
51
+ "action": [
52
+ 7
53
+ ]
54
+ },
55
+ "pretrained_backbone_weights": null,
56
+ "primary_code_loss_weight": 5.0,
57
+ "secondary_code_loss_weight": 0.5,
58
+ "sequentially_select": false,
59
+ "spatial_softmax_num_keypoints": 32,
60
+ "use_group_norm": true,
61
+ "vision_backbone": "resnet18",
62
+ "vqvae_embedding_dim": 256,
63
+ "vqvae_enc_hidden_dim": 128,
64
+ "vqvae_n_embed": 16
65
+ }
config.yaml ADDED
@@ -0,0 +1,210 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ resume: false
2
+ device: cuda
3
+ use_amp: false
4
+ seed: 1
5
+ dataset_repo_id: xarm_holi_demos_lerobot_state_as_action
6
+ video_backend: pyav
7
+ training:
8
+ offline_steps: 250000
9
+ split: train
10
+ num_workers: 4
11
+ batch_size: 32
12
+ eval_freq: -1
13
+ log_freq: 200
14
+ save_checkpoint: true
15
+ save_freq: 25000
16
+ online_steps: 0
17
+ online_rollout_n_episodes: 1
18
+ online_rollout_batch_size: 1
19
+ online_steps_between_rollouts: 1
20
+ online_sampling_ratio: 0.5
21
+ online_env_seed: null
22
+ online_buffer_capacity: null
23
+ online_buffer_seed_size: 0
24
+ do_online_rollout_async: false
25
+ image_transforms:
26
+ enable: false
27
+ max_num_transforms: 3
28
+ random_order: false
29
+ brightness:
30
+ weight: 1
31
+ min_max:
32
+ - 0.8
33
+ - 1.2
34
+ contrast:
35
+ weight: 1
36
+ min_max:
37
+ - 0.8
38
+ - 1.2
39
+ saturation:
40
+ weight: 1
41
+ min_max:
42
+ - 0.5
43
+ - 1.5
44
+ hue:
45
+ weight: 1
46
+ min_max:
47
+ - -0.05
48
+ - 0.05
49
+ sharpness:
50
+ weight: 1
51
+ min_max:
52
+ - 0.8
53
+ - 1.2
54
+ grad_clip_norm: 10
55
+ lr: 0.0001
56
+ lr_scheduler: cosine
57
+ lr_warmup_steps: 500
58
+ adam_betas:
59
+ - 0.95
60
+ - 0.999
61
+ adam_eps: 1.0e-08
62
+ adam_weight_decay: 1.0e-06
63
+ vqvae_lr: 0.001
64
+ n_vqvae_training_steps: 20000
65
+ bet_weight_decay: 0.0002
66
+ bet_learning_rate: 5.5e-05
67
+ bet_betas:
68
+ - 0.9
69
+ - 0.999
70
+ delta_timestamps:
71
+ observation.image.cam_left:
72
+ - -0.13333333333333333
73
+ - -0.1
74
+ - -0.06666666666666667
75
+ - -0.03333333333333333
76
+ - 0.0
77
+ observation.image.cam_right:
78
+ - -0.13333333333333333
79
+ - -0.1
80
+ - -0.06666666666666667
81
+ - -0.03333333333333333
82
+ - 0.0
83
+ observation.image.cam_wrist:
84
+ - -0.13333333333333333
85
+ - -0.1
86
+ - -0.06666666666666667
87
+ - -0.03333333333333333
88
+ - 0.0
89
+ observation.state:
90
+ - -0.13333333333333333
91
+ - -0.1
92
+ - -0.06666666666666667
93
+ - -0.03333333333333333
94
+ - 0.0
95
+ action:
96
+ - -0.13333333333333333
97
+ - -0.1
98
+ - -0.06666666666666667
99
+ - -0.03333333333333333
100
+ - 0.0
101
+ - 0.03333333333333333
102
+ - 0.06666666666666667
103
+ - 0.1
104
+ - 0.13333333333333333
105
+ - 0.16666666666666666
106
+ - 0.2
107
+ - 0.23333333333333334
108
+ - 0.26666666666666666
109
+ - 0.3
110
+ - 0.3333333333333333
111
+ eval:
112
+ n_episodes: 10
113
+ batch_size: 4
114
+ use_async_envs: false
115
+ wandb:
116
+ enable: true
117
+ name: ''
118
+ disable_artifact: false
119
+ project: lerobot
120
+ notes: xarmholi
121
+ fps: 30
122
+ env:
123
+ name: real_world
124
+ task: null
125
+ state_dim: 7
126
+ action_dim: 7
127
+ fps: ${fps}
128
+ override_dataset_stats:
129
+ observation.image.cam_left:
130
+ mean:
131
+ - - - 0.485
132
+ - - - 0.456
133
+ - - - 0.406
134
+ std:
135
+ - - - 0.229
136
+ - - - 0.224
137
+ - - - 0.225
138
+ observation.image.cam_right:
139
+ mean:
140
+ - - - 0.485
141
+ - - - 0.456
142
+ - - - 0.406
143
+ std:
144
+ - - - 0.229
145
+ - - - 0.224
146
+ - - - 0.225
147
+ observation.image.cam_wrist:
148
+ mean:
149
+ - - - 0.485
150
+ - - - 0.456
151
+ - - - 0.406
152
+ std:
153
+ - - - 0.229
154
+ - - - 0.224
155
+ - - - 0.225
156
+ policy:
157
+ name: vqbet
158
+ n_obs_steps: 5
159
+ n_action_pred_token: 7
160
+ action_chunk_size: 5
161
+ input_shapes:
162
+ observation.image.cam_left:
163
+ - 3
164
+ - 96
165
+ - 96
166
+ observation.image.cam_right:
167
+ - 3
168
+ - 96
169
+ - 96
170
+ observation.image.cam_wrist:
171
+ - 3
172
+ - 96
173
+ - 96
174
+ observation.state:
175
+ - ${env.state_dim}
176
+ output_shapes:
177
+ action:
178
+ - ${env.action_dim}
179
+ input_normalization_modes:
180
+ observation.image.cam_left: mean_std
181
+ observation.image.cam_right: mean_std
182
+ observation.image.cam_wrist: mean_std
183
+ observation.state: mean_std
184
+ output_normalization_modes:
185
+ action: mean_std
186
+ vision_backbone: resnet18
187
+ crop_shape:
188
+ - 84
189
+ - 84
190
+ crop_is_random: true
191
+ pretrained_backbone_weights: null
192
+ use_group_norm: true
193
+ spatial_softmax_num_keypoints: 32
194
+ n_vqvae_training_steps: ${training.n_vqvae_training_steps}
195
+ vqvae_n_embed: 16
196
+ vqvae_embedding_dim: 256
197
+ vqvae_enc_hidden_dim: 128
198
+ gpt_block_size: 500
199
+ gpt_input_dim: 512
200
+ gpt_output_dim: 512
201
+ gpt_n_layer: 8
202
+ gpt_n_head: 8
203
+ gpt_hidden_dim: 512
204
+ dropout: 0.1
205
+ mlp_hidden_dim: 1024
206
+ offset_loss_weight: 10000.0
207
+ primary_code_loss_weight: 5.0
208
+ secondary_code_loss_weight: 0.5
209
+ bet_softmax_temperature: 0.1
210
+ sequentially_select: false
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b672c0ed9f105991dc292b08d92f19f188474159440d2ad679bb306b7788642
3
+ size 159830662