Hugging Face's logo

danielsanjosepro
/
cascaded_flow_stack_v2

Robotics
LeRobot
Safetensors
cascaded_flow
Model card Files
xet
Community
danielsanjosepro commited on
Commit
a6d692c
·
verified ·
1 Parent(s): 3a6054b

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +133 -0
  3. model.safetensors +3 -0
  4. train_config.json +243 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: LSY-lab/stack_v2
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: cascaded_flow
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - robotics
9
+ - cascaded_flow
10
+ - lerobot
11
+ ---
12
+
13
+ # Model Card for cascaded_flow
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 +1,133 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "cascaded_flow",
3
+ "n_obs_steps": 2,
4
+ "normalization_mapping": {
5
+ "VISUAL": "MEAN_STD",
6
+ "STATE": "MIN_MAX",
7
+ "ACTION": "MIN_MAX"
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/cascaded_flow_stack_v2",
89
+ "private": null,
90
+ "tags": null,
91
+ "license": null,
92
+ "slow_policy_config": {
93
+ "crop_shape": [
94
+ 224,
95
+ 224
96
+ ],
97
+ "use_separate_rgb_encoder_per_camera": true,
98
+ "horizon": 32,
99
+ "n_action_steps": 16,
100
+ "n_obs_steps": 2,
101
+ "optimizer_lr": 0.0001,
102
+ "training_noise_sampling": "uniform",
103
+ "features_to_exclude": [
104
+ "observation.images.tactile"
105
+ ]
106
+ },
107
+ "fast_policy_config": {
108
+ "crop_shape": [
109
+ 224,
110
+ 224
111
+ ],
112
+ "use_separate_rgb_encoder_per_camera": true,
113
+ "horizon": 8,
114
+ "n_action_steps": 1,
115
+ "n_obs_steps": 2,
116
+ "optimizer_lr": 0.0001,
117
+ "time_distribution": "uniform",
118
+ "flow_ratio": 0.5,
119
+ "features_to_exclude": [
120
+ "observation.images.primary",
121
+ "observation.images.wrist",
122
+ "observation.images.tactile"
123
+ ]
124
+ },
125
+ "action_batch_size": 20,
126
+ "sampling_strategy": "stochastic",
127
+ "sampling_temperature": 1.0,
128
+ "smooth_slow_policy_output": false,
129
+ "apply_slow_policy_output": false,
130
+ "use_image_features_for_fast_policy": true,
131
+ "training_stage": "slow_policy",
132
+ "slow_policy_type": "ditmeanflow"
133
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3e6515fb6620eddf8f6492be041da36527726451786c1b5ddb71165e8122bbe
3
+ size 427947384
train_config.json ADDED
@@ -0,0 +1,243 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "LSY-lab/stack_v2",
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": "cascaded_flow",
70
+ "n_obs_steps": 2,
71
+ "normalization_mapping": {
72
+ "VISUAL": "MEAN_STD",
73
+ "STATE": "MIN_MAX",
74
+ "ACTION": "MIN_MAX"
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/cascaded_flow_stack_v2",
156
+ "private": null,
157
+ "tags": null,
158
+ "license": null,
159
+ "slow_policy_config": {
160
+ "crop_shape": [
161
+ 224,
162
+ 224
163
+ ],
164
+ "use_separate_rgb_encoder_per_camera": true,
165
+ "horizon": 32,
166
+ "n_action_steps": 16,
167
+ "n_obs_steps": 2,
168
+ "optimizer_lr": 0.0001,
169
+ "training_noise_sampling": "uniform",
170
+ "features_to_exclude": [
171
+ "observation.images.tactile"
172
+ ]
173
+ },
174
+ "fast_policy_config": {
175
+ "crop_shape": [
176
+ 224,
177
+ 224
178
+ ],
179
+ "use_separate_rgb_encoder_per_camera": true,
180
+ "horizon": 8,
181
+ "n_action_steps": 1,
182
+ "n_obs_steps": 2,
183
+ "optimizer_lr": 0.0001,
184
+ "time_distribution": "uniform",
185
+ "flow_ratio": 0.5,
186
+ "features_to_exclude": [
187
+ "observation.images.primary",
188
+ "observation.images.wrist",
189
+ "observation.images.tactile"
190
+ ]
191
+ },
192
+ "action_batch_size": 20,
193
+ "sampling_strategy": "stochastic",
194
+ "sampling_temperature": 1.0,
195
+ "smooth_slow_policy_output": false,
196
+ "apply_slow_policy_output": false,
197
+ "use_image_features_for_fast_policy": true,
198
+ "training_stage": "slow_policy",
199
+ "slow_policy_type": "ditmeanflow"
200
+ },
201
+ "output_dir": "outputs/train/2025-12-12/13-36-01_config/train_cascaded_stack.json_slow",
202
+ "job_name": "config/train_cascaded_stack.json_slow",
203
+ "resume": false,
204
+ "seed": 1000,
205
+ "num_workers": 4,
206
+ "batch_size": 128,
207
+ "steps": 20000,
208
+ "eval_freq": 1000,
209
+ "log_freq": 100,
210
+ "save_checkpoint": true,
211
+ "save_freq": 5000,
212
+ "use_policy_training_preset": true,
213
+ "optimizer": {
214
+ "type": "adam",
215
+ "lr": 0.0001,
216
+ "weight_decay": 1e-06,
217
+ "grad_clip_norm": 10.0,
218
+ "betas": [
219
+ 0.95,
220
+ 0.999
221
+ ],
222
+ "eps": 1e-08
223
+ },
224
+ "scheduler": {
225
+ "type": "diffuser",
226
+ "num_warmup_steps": 500,
227
+ "name": "cosine"
228
+ },
229
+ "eval": {
230
+ "n_episodes": 50,
231
+ "batch_size": 50,
232
+ "use_async_envs": false
233
+ },
234
+ "wandb": {
235
+ "enable": true,
236
+ "disable_artifact": false,
237
+ "project": "policies",
238
+ "entity": null,
239
+ "notes": null,
240
+ "run_id": "sxr03v0n",
241
+ "mode": null
242
+ }
243
+ }