Hugging Face's logo

danielsanjosepro
/
dit_meanflow_stack_cake_v1

Robotics
LeRobot
Safetensors
dit_meanflow
Model card Files
xet
Community
danielsanjosepro commited on
Commit
8cd3f79
·
verified ·
1 Parent(s): 3370a2a

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +138 -0
  3. model.safetensors +3 -0
  4. train_config.json +250 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: LSY-lab/stack_cake_v1
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: dit_meanflow
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - dit_meanflow
9
+ - lerobot
10
+ - robotics
11
+ ---
12
+
13
+ # Model Card for dit_meanflow
14
+
15
+ <!-- Provide a quick summary of what the model is/does. -->
16
+
17
+
18
+ _Model type not recognized — please update this template._
19
+
20
+
21
+ This policy has been trained and pushed to the Hub using [LeRobot](https://github.com/huggingface/lerobot).
22
+ See the full documentation at [LeRobot Docs](https://huggingface.co/docs/lerobot/index).
23
+
24
+ ---
25
+
26
+ ## How to Get Started with the Model
27
+
28
+ For a complete walkthrough, see the [training guide](https://huggingface.co/docs/lerobot/il_robots#train-a-policy).
29
+ Below is the short version on how to train and run inference/eval:
30
+
31
+ ### Train from scratch
32
+
33
+ ```bash
34
+ python -m lerobot.scripts.train \
35
+ --dataset.repo_id=${HF_USER}/<dataset> \
36
+ --policy.type=act \
37
+ --output_dir=outputs/train/<desired_policy_repo_id> \
38
+ --job_name=lerobot_training \
39
+ --policy.device=cuda \
40
+ --policy.repo_id=${HF_USER}/<desired_policy_repo_id>
41
+ --wandb.enable=true
42
+ ```
43
+
44
+ *Writes checkpoints to `outputs/train/<desired_policy_repo_id>/checkpoints/`.*
45
+
46
+ ### Evaluate the policy/run inference
47
+
48
+ ```bash
49
+ python -m lerobot.record \
50
+ --robot.type=so100_follower \
51
+ --dataset.repo_id=<hf_user>/eval_<dataset> \
52
+ --policy.path=<hf_user>/<desired_policy_repo_id> \
53
+ --episodes=10
54
+ ```
55
+
56
+ Prefix the dataset repo with **eval\_** and supply `--policy.path` pointing to a local or hub checkpoint.
57
+
58
+ ---
59
+
60
+ ## Model Details
61
+
62
+ * **License:** apache-2.0
config.json ADDED
@@ -0,0 +1,138 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "dit_meanflow",
3
+ "n_obs_steps": 2,
4
+ "normalization_mapping": {
5
+ "VISUAL": "MEAN_STD",
6
+ "STATE": "MIN_MAX",
7
+ "ACTION": "MEAN_STD"
8
+ },
9
+ "input_features": {
10
+ "observation.images.primary": {
11
+ "type": "VISUAL",
12
+ "shape": [
13
+ 3,
14
+ 256,
15
+ 256
16
+ ]
17
+ },
18
+ "observation.images.tactile": {
19
+ "type": "VISUAL",
20
+ "shape": [
21
+ 3,
22
+ 256,
23
+ 256
24
+ ]
25
+ },
26
+ "observation.images.wrist": {
27
+ "type": "VISUAL",
28
+ "shape": [
29
+ 3,
30
+ 256,
31
+ 256
32
+ ]
33
+ },
34
+ "observation.state.cartesian": {
35
+ "type": "STATE",
36
+ "shape": [
37
+ 6
38
+ ]
39
+ },
40
+ "observation.state.gripper": {
41
+ "type": "STATE",
42
+ "shape": [
43
+ 1
44
+ ]
45
+ },
46
+ "observation.state.joints": {
47
+ "type": "STATE",
48
+ "shape": [
49
+ 7
50
+ ]
51
+ },
52
+ "observation.state.sensors_ft_sensor": {
53
+ "type": "STATE",
54
+ "shape": [
55
+ 6
56
+ ]
57
+ },
58
+ "observation.state.sensors_tactile_sensor": {
59
+ "type": "STATE",
60
+ "shape": [
61
+ 15
62
+ ]
63
+ },
64
+ "observation.state.target": {
65
+ "type": "STATE",
66
+ "shape": [
67
+ 6
68
+ ]
69
+ },
70
+ "observation.state": {
71
+ "type": "STATE",
72
+ "shape": [
73
+ 41
74
+ ]
75
+ }
76
+ },
77
+ "output_features": {
78
+ "action": {
79
+ "type": "ACTION",
80
+ "shape": [
81
+ 7
82
+ ]
83
+ }
84
+ },
85
+ "device": "cuda",
86
+ "use_amp": false,
87
+ "push_to_hub": true,
88
+ "repo_id": "danielsanjosepro/dit_meanflow_stack_cake_v1",
89
+ "private": null,
90
+ "tags": null,
91
+ "license": null,
92
+ "horizon": 16,
93
+ "n_action_steps": 8,
94
+ "use_proprioceptive": true,
95
+ "drop_n_last_frames": 7,
96
+ "vision_backbone": "resnet18",
97
+ "crop_shape": null,
98
+ "crop_is_random": false,
99
+ "pretrained_backbone_weights": null,
100
+ "use_group_norm": true,
101
+ "spatial_softmax_num_keypoints": 32,
102
+ "use_separate_rgb_encoder_per_camera": true,
103
+ "frequency_embedding_dim": 256,
104
+ "hidden_dim": 512,
105
+ "num_blocks": 6,
106
+ "num_heads": 16,
107
+ "dropout": 0.1,
108
+ "dim_feedforward": 4096,
109
+ "activation": "gelu",
110
+ "training_noise_sampling": "uniform",
111
+ "clip_sample": true,
112
+ "clip_sample_range": 1.0,
113
+ "num_inference_steps": 100,
114
+ "do_mask_loss_for_padding": false,
115
+ "optimizer_lr": 0.0001,
116
+ "optimizer_betas": [
117
+ 0.95,
118
+ 0.999
119
+ ],
120
+ "optimizer_eps": 1e-08,
121
+ "optimizer_weight_decay": 1e-06,
122
+ "scheduler_name": "cosine",
123
+ "scheduler_warmup_steps": 500,
124
+ "use_meanflow": true,
125
+ "do_multi_step_sampling": true,
126
+ "inference_timesteps": 100,
127
+ "time_distribution": "uniform",
128
+ "log_norm_mu": -0.4,
129
+ "log_norm_sigma": 1.0,
130
+ "use_autograd_functional_jvp": false,
131
+ "use_adaptative_loss": true,
132
+ "flow_ratio": 0.5,
133
+ "cfg_prob": 0.1,
134
+ "cfg_omega": 2.0,
135
+ "cfg_w": 2.0,
136
+ "cfg_kappa": 0.0,
137
+ "optimizer_grad_clip_norm": 0.5
138
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fca7aef22ac61fcbb36b7f249eee873121604074a82a150f1aee0dacf760dc98
3
+ size 304553516
train_config.json ADDED
@@ -0,0 +1,250 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "LSY-lab/stack_cake_v1",
4
+ "root": null,
5
+ "episodes": null,
6
+ "image_transforms": {
7
+ "enable": false,
8
+ "max_num_transforms": 3,
9
+ "random_order": false,
10
+ "tfs": {
11
+ "brightness": {
12
+ "weight": 1.0,
13
+ "type": "ColorJitter",
14
+ "kwargs": {
15
+ "brightness": [
16
+ 0.8,
17
+ 1.2
18
+ ]
19
+ }
20
+ },
21
+ "contrast": {
22
+ "weight": 1.0,
23
+ "type": "ColorJitter",
24
+ "kwargs": {
25
+ "contrast": [
26
+ 0.8,
27
+ 1.2
28
+ ]
29
+ }
30
+ },
31
+ "saturation": {
32
+ "weight": 1.0,
33
+ "type": "ColorJitter",
34
+ "kwargs": {
35
+ "saturation": [
36
+ 0.5,
37
+ 1.5
38
+ ]
39
+ }
40
+ },
41
+ "hue": {
42
+ "weight": 1.0,
43
+ "type": "ColorJitter",
44
+ "kwargs": {
45
+ "hue": [
46
+ -0.05,
47
+ 0.05
48
+ ]
49
+ }
50
+ },
51
+ "sharpness": {
52
+ "weight": 1.0,
53
+ "type": "SharpnessJitter",
54
+ "kwargs": {
55
+ "sharpness": [
56
+ 0.5,
57
+ 1.5
58
+ ]
59
+ }
60
+ }
61
+ }
62
+ },
63
+ "revision": null,
64
+ "use_imagenet_stats": true,
65
+ "video_backend": "torchcodec"
66
+ },
67
+ "env": null,
68
+ "policy": {
69
+ "type": "dit_meanflow",
70
+ "n_obs_steps": 2,
71
+ "normalization_mapping": {
72
+ "VISUAL": "MEAN_STD",
73
+ "STATE": "MIN_MAX",
74
+ "ACTION": "MEAN_STD"
75
+ },
76
+ "input_features": {
77
+ "observation.images.primary": {
78
+ "type": "VISUAL",
79
+ "shape": [
80
+ 3,
81
+ 256,
82
+ 256
83
+ ]
84
+ },
85
+ "observation.images.tactile": {
86
+ "type": "VISUAL",
87
+ "shape": [
88
+ 3,
89
+ 256,
90
+ 256
91
+ ]
92
+ },
93
+ "observation.images.wrist": {
94
+ "type": "VISUAL",
95
+ "shape": [
96
+ 3,
97
+ 256,
98
+ 256
99
+ ]
100
+ },
101
+ "observation.state.cartesian": {
102
+ "type": "STATE",
103
+ "shape": [
104
+ 6
105
+ ]
106
+ },
107
+ "observation.state.gripper": {
108
+ "type": "STATE",
109
+ "shape": [
110
+ 1
111
+ ]
112
+ },
113
+ "observation.state.joints": {
114
+ "type": "STATE",
115
+ "shape": [
116
+ 7
117
+ ]
118
+ },
119
+ "observation.state.sensors_ft_sensor": {
120
+ "type": "STATE",
121
+ "shape": [
122
+ 6
123
+ ]
124
+ },
125
+ "observation.state.sensors_tactile_sensor": {
126
+ "type": "STATE",
127
+ "shape": [
128
+ 15
129
+ ]
130
+ },
131
+ "observation.state.target": {
132
+ "type": "STATE",
133
+ "shape": [
134
+ 6
135
+ ]
136
+ },
137
+ "observation.state": {
138
+ "type": "STATE",
139
+ "shape": [
140
+ 41
141
+ ]
142
+ }
143
+ },
144
+ "output_features": {
145
+ "action": {
146
+ "type": "ACTION",
147
+ "shape": [
148
+ 7
149
+ ]
150
+ }
151
+ },
152
+ "device": "cuda",
153
+ "use_amp": false,
154
+ "push_to_hub": true,
155
+ "repo_id": "danielsanjosepro/dit_meanflow_stack_cake_v1",
156
+ "private": null,
157
+ "tags": null,
158
+ "license": null,
159
+ "horizon": 16,
160
+ "n_action_steps": 8,
161
+ "use_proprioceptive": true,
162
+ "drop_n_last_frames": 7,
163
+ "vision_backbone": "resnet18",
164
+ "crop_shape": null,
165
+ "crop_is_random": false,
166
+ "pretrained_backbone_weights": null,
167
+ "use_group_norm": true,
168
+ "spatial_softmax_num_keypoints": 32,
169
+ "use_separate_rgb_encoder_per_camera": true,
170
+ "frequency_embedding_dim": 256,
171
+ "hidden_dim": 512,
172
+ "num_blocks": 6,
173
+ "num_heads": 16,
174
+ "dropout": 0.1,
175
+ "dim_feedforward": 4096,
176
+ "activation": "gelu",
177
+ "training_noise_sampling": "uniform",
178
+ "clip_sample": true,
179
+ "clip_sample_range": 1.0,
180
+ "num_inference_steps": 100,
181
+ "do_mask_loss_for_padding": false,
182
+ "optimizer_lr": 0.0001,
183
+ "optimizer_betas": [
184
+ 0.95,
185
+ 0.999
186
+ ],
187
+ "optimizer_eps": 1e-08,
188
+ "optimizer_weight_decay": 1e-06,
189
+ "scheduler_name": "cosine",
190
+ "scheduler_warmup_steps": 500,
191
+ "use_meanflow": true,
192
+ "do_multi_step_sampling": true,
193
+ "inference_timesteps": 100,
194
+ "time_distribution": "uniform",
195
+ "log_norm_mu": -0.4,
196
+ "log_norm_sigma": 1.0,
197
+ "use_autograd_functional_jvp": false,
198
+ "use_adaptative_loss": true,
199
+ "flow_ratio": 0.5,
200
+ "cfg_prob": 0.1,
201
+ "cfg_omega": 2.0,
202
+ "cfg_w": 2.0,
203
+ "cfg_kappa": 0.0,
204
+ "optimizer_grad_clip_norm": 0.5
205
+ },
206
+ "output_dir": "outputs/train/2025-11-15/18-06-51_dit_meanflow_stack_cake_v2",
207
+ "job_name": "dit_meanflow_stack_cake_v2",
208
+ "resume": false,
209
+ "seed": 1000,
210
+ "num_workers": 4,
211
+ "batch_size": 64,
212
+ "steps": 20000,
213
+ "eval_freq": 20000,
214
+ "log_freq": 100,
215
+ "save_checkpoint": true,
216
+ "save_freq": 5000,
217
+ "use_policy_training_preset": true,
218
+ "optimizer": {
219
+ "type": "adam",
220
+ "lr": 0.0001,
221
+ "weight_decay": 1e-06,
222
+ "grad_clip_norm": 10.0,
223
+ "betas": [
224
+ 0.95,
225
+ 0.999
226
+ ],
227
+ "eps": 1e-08
228
+ },
229
+ "scheduler": {
230
+ "type": "diffuser",
231
+ "num_warmup_steps": 500,
232
+ "name": "cosine"
233
+ },
234
+ "eval": {
235
+ "n_episodes": 50,
236
+ "batch_size": 50,
237
+ "use_async_envs": false,
238
+ "max_episodes_rendered": 4
239
+ },
240
+ "wandb": {
241
+ "enable": true,
242
+ "disable_artifact": false,
243
+ "project": "lerobot",
244
+ "entity": null,
245
+ "notes": null,
246
+ "run_id": "adtf9082",
247
+ "mode": null
248
+ },
249
+ "eval_with_dataset": "LSY-lab/stack_cake_v1"
250
+ }