Hugging Face's logo

lerobot
/
diffusion_pusht_keypoints

Robotics
Transformers
Safetensors
Inference Endpoints
Model card Files Community
2
imstevenpmwork HF staff commited on
Commit
286e5ea
·
verified ·
1 Parent(s): 58570fc

Upload folder using huggingface_hub

Browse files
Files changed (2) hide show
  1. config.json +54 -51
  2. train_config.json +194 -0
config.json CHANGED
@@ -1,53 +1,56 @@
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": 128,
13
- "do_mask_loss_for_padding": false,
14
- "down_dims": [
15
- 512,
16
- 1024,
17
- 2048
18
- ],
19
- "horizon": 16,
20
- "input_normalization_modes": {
21
- "observation.environment_state": "min_max",
22
- "observation.state": "min_max"
23
- },
24
- "input_shapes": {
25
- "observation.environment_state": [
26
- 16
27
- ],
28
- "observation.state": [
29
- 2
30
- ]
31
- },
32
- "kernel_size": 5,
33
- "n_action_steps": 8,
34
- "n_groups": 8,
35
- "n_obs_steps": 2,
36
- "noise_scheduler_type": "DDIM",
37
- "num_inference_steps": 10,
38
- "num_train_timesteps": 100,
39
- "output_normalization_modes": {
40
- "action": "min_max"
41
- },
42
- "output_shapes": {
43
- "action": [
44
- 2
45
- ]
46
- },
47
- "prediction_type": "epsilon",
48
- "pretrained_backbone_weights": null,
49
- "spatial_softmax_num_keypoints": 32,
50
- "use_film_scale_modulation": true,
51
- "use_group_norm": true,
52
- "vision_backbone": "resnet18"
 
 
 
53
  }
 
1
  {
2
+ "type": "act",
3
+ "n_obs_steps": 1,
4
+ "device": "cuda",
5
+ "use_amp": false,
6
+ "normalization_mapping": {
7
+ "VISUAL": "MEAN_STD",
8
+ "STATE": "MEAN_STD",
9
+ "ACTION": "MEAN_STD"
10
+ },
11
+ "input_features": {
12
+ "observation.images.top": {
13
+ "type": "VISUAL",
14
+ "shape": [
15
+ 3,
16
+ 480,
17
+ 640
18
+ ]
19
+ },
20
+ "observation.state": {
21
+ "type": "STATE",
22
+ "shape": [
23
+ 14
24
+ ]
25
+ }
26
+ },
27
+ "output_features": {
28
+ "action": {
29
+ "type": "ACTION",
30
+ "shape": [
31
+ 14
32
+ ]
33
+ }
34
+ },
35
+ "chunk_size": 100,
36
+ "n_action_steps": 100,
37
+ "vision_backbone": "resnet18",
38
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
39
+ "replace_final_stride_with_dilation": false,
40
+ "pre_norm": false,
41
+ "dim_model": 512,
42
+ "n_heads": 8,
43
+ "dim_feedforward": 3200,
44
+ "feedforward_activation": "relu",
45
+ "n_encoder_layers": 4,
46
+ "n_decoder_layers": 1,
47
+ "use_vae": true,
48
+ "latent_dim": 32,
49
+ "n_vae_encoder_layers": 4,
50
+ "temporal_ensemble_coeff": null,
51
+ "dropout": 0.1,
52
+ "kl_weight": 10.0,
53
+ "optimizer_lr": 1e-05,
54
+ "optimizer_weight_decay": 0.0001,
55
+ "optimizer_lr_backbone": 1e-05
56
  }
train_config.json ADDED
@@ -0,0 +1,194 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "lerobot/aloha_sim_insertion_human",
4
+ "episodes": null,
5
+ "image_transforms": {
6
+ "enable": false,
7
+ "max_num_transforms": 3,
8
+ "random_order": false,
9
+ "tfs": {
10
+ "brightness": {
11
+ "weight": 1.0,
12
+ "type": "ColorJitter",
13
+ "kwargs": {
14
+ "brightness": [
15
+ 0.8,
16
+ 1.2
17
+ ]
18
+ }
19
+ },
20
+ "contrast": {
21
+ "weight": 1.0,
22
+ "type": "ColorJitter",
23
+ "kwargs": {
24
+ "contrast": [
25
+ 0.8,
26
+ 1.2
27
+ ]
28
+ }
29
+ },
30
+ "saturation": {
31
+ "weight": 1.0,
32
+ "type": "ColorJitter",
33
+ "kwargs": {
34
+ "saturation": [
35
+ 0.5,
36
+ 1.5
37
+ ]
38
+ }
39
+ },
40
+ "hue": {
41
+ "weight": 1.0,
42
+ "type": "ColorJitter",
43
+ "kwargs": {
44
+ "hue": [
45
+ -0.05,
46
+ 0.05
47
+ ]
48
+ }
49
+ },
50
+ "sharpness": {
51
+ "weight": 1.0,
52
+ "type": "SharpnessJitter",
53
+ "kwargs": {
54
+ "sharpness": [
55
+ 0.5,
56
+ 1.5
57
+ ]
58
+ }
59
+ }
60
+ }
61
+ },
62
+ "local_files_only": false,
63
+ "use_imagenet_stats": true,
64
+ "video_backend": "pyav"
65
+ },
66
+ "env": {
67
+ "type": "aloha",
68
+ "task": "AlohaInsertion-v0",
69
+ "fps": 50,
70
+ "features": {
71
+ "action": {
72
+ "type": "ACTION",
73
+ "shape": [
74
+ 14
75
+ ]
76
+ },
77
+ "agent_pos": {
78
+ "type": "STATE",
79
+ "shape": [
80
+ 14
81
+ ]
82
+ },
83
+ "pixels/top": {
84
+ "type": "VISUAL",
85
+ "shape": [
86
+ 480,
87
+ 640,
88
+ 3
89
+ ]
90
+ }
91
+ },
92
+ "features_map": {
93
+ "action": "action",
94
+ "agent_pos": "observation.state",
95
+ "top": "observation.image.top",
96
+ "pixels/top": "observation.images.top"
97
+ },
98
+ "episode_length": 400,
99
+ "obs_type": "pixels_agent_pos",
100
+ "render_mode": "rgb_array"
101
+ },
102
+ "policy": {
103
+ "type": "act",
104
+ "n_obs_steps": 1,
105
+ "device": "cuda",
106
+ "use_amp": false,
107
+ "normalization_mapping": {
108
+ "VISUAL": "MEAN_STD",
109
+ "STATE": "MEAN_STD",
110
+ "ACTION": "MEAN_STD"
111
+ },
112
+ "input_features": {
113
+ "observation.images.top": {
114
+ "type": "VISUAL",
115
+ "shape": [
116
+ 3,
117
+ 480,
118
+ 640
119
+ ]
120
+ },
121
+ "observation.state": {
122
+ "type": "STATE",
123
+ "shape": [
124
+ 14
125
+ ]
126
+ }
127
+ },
128
+ "output_features": {
129
+ "action": {
130
+ "type": "ACTION",
131
+ "shape": [
132
+ 14
133
+ ]
134
+ }
135
+ },
136
+ "chunk_size": 100,
137
+ "n_action_steps": 100,
138
+ "vision_backbone": "resnet18",
139
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
140
+ "replace_final_stride_with_dilation": false,
141
+ "pre_norm": false,
142
+ "dim_model": 512,
143
+ "n_heads": 8,
144
+ "dim_feedforward": 3200,
145
+ "feedforward_activation": "relu",
146
+ "n_encoder_layers": 4,
147
+ "n_decoder_layers": 1,
148
+ "use_vae": true,
149
+ "latent_dim": 32,
150
+ "n_vae_encoder_layers": 4,
151
+ "temporal_ensemble_coeff": null,
152
+ "dropout": 0.1,
153
+ "kl_weight": 10.0,
154
+ "optimizer_lr": 1e-05,
155
+ "optimizer_weight_decay": 0.0001,
156
+ "optimizer_lr_backbone": 1e-05
157
+ },
158
+ "output_dir": "outputs/train/2025-01-25/17-37-25_aloha_act",
159
+ "job_name": "aloha_act",
160
+ "resume": false,
161
+ "seed": 1000,
162
+ "num_workers": 4,
163
+ "batch_size": 8,
164
+ "steps": 100000,
165
+ "eval_freq": 20000,
166
+ "log_freq": 200,
167
+ "save_checkpoint": true,
168
+ "save_freq": 20000,
169
+ "use_policy_training_preset": true,
170
+ "optimizer": {
171
+ "type": "adamw",
172
+ "lr": 1e-05,
173
+ "betas": [
174
+ 0.9,
175
+ 0.999
176
+ ],
177
+ "eps": 1e-08,
178
+ "weight_decay": 0.0001,
179
+ "grad_clip_norm": 10.0
180
+ },
181
+ "scheduler": null,
182
+ "eval": {
183
+ "n_episodes": 50,
184
+ "batch_size": 50,
185
+ "use_async_envs": false
186
+ },
187
+ "wandb": {
188
+ "enable": true,
189
+ "disable_artifact": false,
190
+ "project": "lerobot",
191
+ "entity": null,
192
+ "notes": null
193
+ }
194
+ }