Hugging Face's logo

aliberts
/
diffusion_test

Robotics
LeRobot
Safetensors
diffusion-policy
model_hub_mixin
pytorch_model_hub_mixin
Model card Files
xet
Community
aliberts commited on
Commit
59675ce
·
verified ·
1 Parent(s): dc6c9b3

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. README.md +12 -0
  2. config.json +74 -0
  3. model.safetensors +3 -0
  4. train_config.json +236 -0
README.md ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: lerobot
3
+ tags:
4
+ - diffusion-policy
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 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "beta_end": 0.02,
3
+ "beta_schedule": "squaredcos_cap_v2",
4
+ "beta_start": 0.0001,
5
+ "clip_sample": true,
6
+ "clip_sample_range": 1.0,
7
+ "crop_is_random": true,
8
+ "crop_shape": [
9
+ 84,
10
+ 84
11
+ ],
12
+ "diffusion_step_embed_dim": 32,
13
+ "do_mask_loss_for_padding": false,
14
+ "down_dims": [
15
+ 64,
16
+ 128,
17
+ 256
18
+ ],
19
+ "drop_n_last_frames": 7,
20
+ "horizon": 16,
21
+ "input_features": {
22
+ "observation.image": {
23
+ "shape": [
24
+ 3,
25
+ 96,
26
+ 96
27
+ ],
28
+ "type": "VISUAL"
29
+ },
30
+ "observation.state": {
31
+ "shape": [
32
+ 2
33
+ ],
34
+ "type": "STATE"
35
+ }
36
+ },
37
+ "kernel_size": 5,
38
+ "n_action_steps": 8,
39
+ "n_groups": 8,
40
+ "n_obs_steps": 2,
41
+ "noise_scheduler_type": "DDPM",
42
+ "normalization_mapping": {
43
+ "ACTION": "MIN_MAX",
44
+ "STATE": "MIN_MAX",
45
+ "VISUAL": "MEAN_STD"
46
+ },
47
+ "num_inference_steps": 10,
48
+ "num_train_timesteps": 100,
49
+ "optimizer_betas": [
50
+ 0.95,
51
+ 0.999
52
+ ],
53
+ "optimizer_eps": 1e-08,
54
+ "optimizer_lr": 0.0001,
55
+ "optimizer_weight_decay": 1e-06,
56
+ "output_features": {
57
+ "action": {
58
+ "shape": [
59
+ 2
60
+ ],
61
+ "type": "ACTION"
62
+ }
63
+ },
64
+ "prediction_type": "epsilon",
65
+ "pretrained_backbone_weights": null,
66
+ "scheduler_name": "cosine",
67
+ "scheduler_warmup_steps": 500,
68
+ "spatial_softmax_num_keypoints": 32,
69
+ "type": "diffusion",
70
+ "use_film_scale_modulation": true,
71
+ "use_group_norm": true,
72
+ "use_separate_rgb_encoder_per_camera": false,
73
+ "vision_backbone": "resnet18"
74
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3dcf5ed4f84a62c2b88aba90e754e8582b8c2a825a02b39b96b9a4801cfc07c5
3
+ size 62501608
train_config.json ADDED
@@ -0,0 +1,236 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "lerobot/pusht",
4
+ "episodes": [
5
+ 0
6
+ ],
7
+ "image_transforms": {
8
+ "enable": true,
9
+ "max_num_transforms": 3,
10
+ "random_order": false,
11
+ "tfs": {
12
+ "brightness": {
13
+ "weight": 1.0,
14
+ "type": "ColorJitter",
15
+ "kwargs": {
16
+ "brightness": [
17
+ 0.8,
18
+ 1.2
19
+ ]
20
+ }
21
+ },
22
+ "contrast": {
23
+ "weight": 1.0,
24
+ "type": "ColorJitter",
25
+ "kwargs": {
26
+ "contrast": [
27
+ 0.8,
28
+ 1.2
29
+ ]
30
+ }
31
+ },
32
+ "saturation": {
33
+ "weight": 1.0,
34
+ "type": "ColorJitter",
35
+ "kwargs": {
36
+ "saturation": [
37
+ 0.5,
38
+ 1.5
39
+ ]
40
+ }
41
+ },
42
+ "hue": {
43
+ "weight": 1.0,
44
+ "type": "ColorJitter",
45
+ "kwargs": {
46
+ "hue": [
47
+ -0.05,
48
+ 0.05
49
+ ]
50
+ }
51
+ },
52
+ "sharpness": {
53
+ "weight": 1.0,
54
+ "type": "SharpnessJitter",
55
+ "kwargs": {
56
+ "sharpness": [
57
+ 0.5,
58
+ 1.5
59
+ ]
60
+ }
61
+ }
62
+ }
63
+ },
64
+ "local_files_only": false,
65
+ "use_imagenet_stats": true,
66
+ "video_backend": "pyav"
67
+ },
68
+ "env": {
69
+ "type": "pusht",
70
+ "n_envs": null,
71
+ "task": "PushT-v0",
72
+ "fps": 10,
73
+ "features": {
74
+ "action": {
75
+ "type": "ACTION",
76
+ "shape": [
77
+ 2
78
+ ]
79
+ },
80
+ "agent_pos": {
81
+ "type": "STATE",
82
+ "shape": [
83
+ 2
84
+ ]
85
+ },
86
+ "pixels": {
87
+ "type": "VISUAL",
88
+ "shape": [
89
+ 384,
90
+ 384,
91
+ 3
92
+ ]
93
+ }
94
+ },
95
+ "features_map": {
96
+ "action": "action",
97
+ "agent_pos": "observation.state",
98
+ "environment_state": "observation.environment_state",
99
+ "pixels": "observation.image"
100
+ },
101
+ "episode_length": 5,
102
+ "obs_type": "pixels_agent_pos",
103
+ "render_mode": "rgb_array",
104
+ "visualization_width": 384,
105
+ "visualization_height": 384
106
+ },
107
+ "policy": {
108
+ "type": "diffusion",
109
+ "n_obs_steps": 2,
110
+ "normalization_mapping": {
111
+ "VISUAL": "MEAN_STD",
112
+ "STATE": "MIN_MAX",
113
+ "ACTION": "MIN_MAX"
114
+ },
115
+ "input_features": {
116
+ "observation.image": {
117
+ "type": "VISUAL",
118
+ "shape": [
119
+ 3,
120
+ 96,
121
+ 96
122
+ ]
123
+ },
124
+ "observation.state": {
125
+ "type": "STATE",
126
+ "shape": [
127
+ 2
128
+ ]
129
+ }
130
+ },
131
+ "output_features": {
132
+ "action": {
133
+ "type": "ACTION",
134
+ "shape": [
135
+ 2
136
+ ]
137
+ }
138
+ },
139
+ "horizon": 16,
140
+ "n_action_steps": 8,
141
+ "drop_n_last_frames": 7,
142
+ "vision_backbone": "resnet18",
143
+ "crop_shape": [
144
+ 84,
145
+ 84
146
+ ],
147
+ "crop_is_random": true,
148
+ "pretrained_backbone_weights": null,
149
+ "use_group_norm": true,
150
+ "spatial_softmax_num_keypoints": 32,
151
+ "use_separate_rgb_encoder_per_camera": false,
152
+ "down_dims": [
153
+ 64,
154
+ 128,
155
+ 256
156
+ ],
157
+ "kernel_size": 5,
158
+ "n_groups": 8,
159
+ "diffusion_step_embed_dim": 32,
160
+ "use_film_scale_modulation": true,
161
+ "noise_scheduler_type": "DDPM",
162
+ "num_train_timesteps": 100,
163
+ "beta_schedule": "squaredcos_cap_v2",
164
+ "beta_start": 0.0001,
165
+ "beta_end": 0.02,
166
+ "prediction_type": "epsilon",
167
+ "clip_sample": true,
168
+ "clip_sample_range": 1.0,
169
+ "num_inference_steps": 10,
170
+ "do_mask_loss_for_padding": false,
171
+ "optimizer_lr": 0.0001,
172
+ "optimizer_betas": [
173
+ 0.95,
174
+ 0.999
175
+ ],
176
+ "optimizer_eps": 1e-08,
177
+ "optimizer_weight_decay": 1e-06,
178
+ "scheduler_name": "cosine",
179
+ "scheduler_warmup_steps": 500
180
+ },
181
+ "output_dir": "outputs/train/pusht_diffusion",
182
+ "job_name": "pusht_diffusion",
183
+ "resume": false,
184
+ "device": "cpu",
185
+ "use_amp": false,
186
+ "seed": 1000,
187
+ "num_workers": 4,
188
+ "batch_size": 2,
189
+ "eval_freq": 20000,
190
+ "log_freq": 1,
191
+ "save_checkpoint": true,
192
+ "save_freq": 2,
193
+ "offline": {
194
+ "steps": 2
195
+ },
196
+ "online": {
197
+ "steps": 0,
198
+ "rollout_n_episodes": 1,
199
+ "rollout_batch_size": 1,
200
+ "steps_between_rollouts": null,
201
+ "sampling_ratio": 0.5,
202
+ "env_seed": null,
203
+ "buffer_capacity": null,
204
+ "buffer_seed_size": 0,
205
+ "do_rollout_async": false
206
+ },
207
+ "use_policy_training_preset": true,
208
+ "optimizer": {
209
+ "type": "adam",
210
+ "lr": 0.0001,
211
+ "betas": [
212
+ 0.95,
213
+ 0.999
214
+ ],
215
+ "eps": 1e-08,
216
+ "weight_decay": 1e-06,
217
+ "grad_clip_norm": 10.0
218
+ },
219
+ "scheduler": {
220
+ "type": "diffuser",
221
+ "num_warmup_steps": 500,
222
+ "name": "cosine"
223
+ },
224
+ "eval": {
225
+ "n_episodes": 1,
226
+ "batch_size": 1,
227
+ "use_async_envs": false
228
+ },
229
+ "wandb": {
230
+ "enable": false,
231
+ "disable_artifact": false,
232
+ "project": "lerobot",
233
+ "entity": null,
234
+ "notes": null
235
+ }
236
+ }