Hugging Face's logo

serialexperimentsleon
/
train_diffusion_pusht_keypoints

Robotics
Safetensors
lerobot
diffusion-policy
model_hub_mixin
pytorch_model_hub_mixin
Model card Files Community
serialexperimentsleon commited on
Commit
c498b25
·
verified ·
1 Parent(s): 897032f

Upload folder using huggingface_hub

Browse files
Files changed (4) hide show
  1. README.md +12 -0
  2. config.json +54 -0
  3. config.yaml +153 -0
  4. model.safetensors +3 -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 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ 256,
16
+ 512,
17
+ 1024
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
+ "use_separate_rgb_encoder_per_camera": false,
53
+ "vision_backbone": "resnet18"
54
+ }
config.yaml ADDED
@@ -0,0 +1,153 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ resume: false
2
+ device: cuda
3
+ use_amp: true
4
+ seed: 100000
5
+ dataset_repo_id: lerobot/pusht_keypoints
6
+ video_backend: pyav
7
+ training:
8
+ offline_steps: 200000
9
+ num_workers: 4
10
+ batch_size: 64
11
+ eval_freq: 20000
12
+ log_freq: 50
13
+ save_checkpoint: true
14
+ save_freq: 20000
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
+ grad_clip_norm: 10
54
+ lr: 0.0001
55
+ lr_scheduler: cosine
56
+ lr_warmup_steps: 500
57
+ adam_betas:
58
+ - 0.95
59
+ - 0.999
60
+ adam_eps: 1.0e-08
61
+ adam_weight_decay: 1.0e-06
62
+ delta_timestamps:
63
+ observation.environment_state:
64
+ - -0.1
65
+ - 0.0
66
+ observation.state:
67
+ - -0.1
68
+ - 0.0
69
+ action:
70
+ - -0.1
71
+ - 0.0
72
+ - 0.1
73
+ - 0.2
74
+ - 0.3
75
+ - 0.4
76
+ - 0.5
77
+ - 0.6
78
+ - 0.7
79
+ - 0.8
80
+ - 0.9
81
+ - 1.0
82
+ - 1.1
83
+ - 1.2
84
+ - 1.3
85
+ - 1.4
86
+ drop_n_last_frames: 7
87
+ eval:
88
+ n_episodes: 50
89
+ batch_size: 50
90
+ use_async_envs: false
91
+ wandb:
92
+ enable: true
93
+ disable_artifact: false
94
+ project: lerobot
95
+ notes: ''
96
+ fps: 10
97
+ env:
98
+ name: pusht
99
+ task: PushT-v0
100
+ image_size: 96
101
+ state_dim: 2
102
+ action_dim: 2
103
+ fps: ${fps}
104
+ episode_length: 300
105
+ gym:
106
+ obs_type: environment_state_agent_pos
107
+ render_mode: rgb_array
108
+ visualization_width: 384
109
+ visualization_height: 384
110
+ policy:
111
+ name: diffusion
112
+ n_obs_steps: 2
113
+ horizon: 16
114
+ n_action_steps: 8
115
+ input_shapes:
116
+ observation.environment_state:
117
+ - 16
118
+ observation.state:
119
+ - ${env.state_dim}
120
+ output_shapes:
121
+ action:
122
+ - ${env.action_dim}
123
+ input_normalization_modes:
124
+ observation.environment_state: min_max
125
+ observation.state: min_max
126
+ output_normalization_modes:
127
+ action: min_max
128
+ vision_backbone: resnet18
129
+ crop_shape:
130
+ - 84
131
+ - 84
132
+ crop_is_random: true
133
+ pretrained_backbone_weights: null
134
+ use_group_norm: true
135
+ spatial_softmax_num_keypoints: 32
136
+ down_dims:
137
+ - 256
138
+ - 512
139
+ - 1024
140
+ kernel_size: 5
141
+ n_groups: 8
142
+ diffusion_step_embed_dim: 128
143
+ use_film_scale_modulation: true
144
+ noise_scheduler_type: DDIM
145
+ num_train_timesteps: 100
146
+ beta_schedule: squaredcos_cap_v2
147
+ beta_start: 0.0001
148
+ beta_end: 0.02
149
+ prediction_type: epsilon
150
+ clip_sample: true
151
+ clip_sample_range: 1.0
152
+ num_inference_steps: 10
153
+ do_mask_loss_for_padding: false
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7e15dca6416adbe8ace5433e97c67e91abd92a55ecfc4df16a9e871d26f8c9fe
3
+ size 254007080