kelvinzhaozg commited on
Commit
b1fbf2a
·
verified ·
1 Parent(s): bf52859

Upload policy weights, train config and readme

Browse files
Files changed (4) hide show
  1. README.md +62 -0
  2. config.json +2 -2
  3. model.safetensors +2 -2
  4. train_config.json +203 -0
README.md ADDED
@@ -0,0 +1,62 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ datasets: kelvinzhaozg/arx_dual_carpet_separation_lerobot
3
+ library_name: lerobot
4
+ license: apache-2.0
5
+ model_name: flow_matching_dit
6
+ pipeline_tag: robotics
7
+ tags:
8
+ - robotics
9
+ - lerobot
10
+ - flow_matching_dit
11
+ ---
12
+
13
+ # Model Card for flow_matching_dit
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
+ lerobot-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
+ 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 CHANGED
@@ -48,7 +48,7 @@
48
  },
49
  "device": "cuda",
50
  "use_amp": false,
51
- "push_to_hub": false,
52
  "repo_id": "kelvinzhaozg/flow_matching_dit_arx_dual_carpet_separation",
53
  "private": null,
54
  "tags": null,
@@ -86,7 +86,7 @@
86
  "scheduler_warmup_steps": 500,
87
  "flow_matching_type": "CondOT",
88
  "num_integration_steps": 50,
89
- "fm_time_embed_dim": 256,
90
  "fm_min_period": 0.004,
91
  "fm_max_period": 4.0,
92
  "ode_solver_method": "midpoint"
 
48
  },
49
  "device": "cuda",
50
  "use_amp": false,
51
+ "push_to_hub": true,
52
  "repo_id": "kelvinzhaozg/flow_matching_dit_arx_dual_carpet_separation",
53
  "private": null,
54
  "tags": null,
 
86
  "scheduler_warmup_steps": 500,
87
  "flow_matching_type": "CondOT",
88
  "num_integration_steps": 50,
89
+ "fm_time_embed_dim": 128,
90
  "fm_min_period": 0.004,
91
  "fm_max_period": 4.0,
92
  "ode_solver_method": "midpoint"
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:4fa844f9c0528f0bf3480fd1c65f94667c7fb24b576583e5111fc99610874980
3
- size 889046856
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:290fa5b858f3694a20444e16c6515d9e62e917eebfa5a9fa32c1cd71c5dbd2c0
3
+ size 832029504
train_config.json ADDED
@@ -0,0 +1,203 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "kelvinzhaozg/arx_dual_carpet_separation_lerobot",
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": "pyav"
66
+ },
67
+ "env": null,
68
+ "policy": {
69
+ "type": "flow_matching_dit",
70
+ "n_obs_steps": 1,
71
+ "normalization_mapping": {
72
+ "VISUAL": "MEAN_STD",
73
+ "STATE": "MIN_MAX",
74
+ "ACTION": "MIN_MAX"
75
+ },
76
+ "input_features": {
77
+ "observation.images.base_image": {
78
+ "type": "VISUAL",
79
+ "shape": [
80
+ 3,
81
+ 240,
82
+ 424
83
+ ]
84
+ },
85
+ "observation.images.right_wrist_image": {
86
+ "type": "VISUAL",
87
+ "shape": [
88
+ 3,
89
+ 240,
90
+ 424
91
+ ]
92
+ },
93
+ "observation.images.left_wrist_image": {
94
+ "type": "VISUAL",
95
+ "shape": [
96
+ 3,
97
+ 240,
98
+ 424
99
+ ]
100
+ },
101
+ "observation.state": {
102
+ "type": "STATE",
103
+ "shape": [
104
+ 14
105
+ ]
106
+ }
107
+ },
108
+ "output_features": {
109
+ "action": {
110
+ "type": "ACTION",
111
+ "shape": [
112
+ 14
113
+ ]
114
+ }
115
+ },
116
+ "device": "cuda",
117
+ "use_amp": false,
118
+ "push_to_hub": true,
119
+ "repo_id": "kelvinzhaozg/flow_matching_dit_arx_dual_carpet_separation",
120
+ "private": null,
121
+ "tags": null,
122
+ "license": null,
123
+ "horizon": 64,
124
+ "n_action_steps": 32,
125
+ "drop_n_last_frames": 7,
126
+ "vision_backbone": "resnet18",
127
+ "crop_shape": [
128
+ 224,
129
+ 224
130
+ ],
131
+ "crop_is_random": true,
132
+ "pretrained_backbone_weights": null,
133
+ "use_group_norm": true,
134
+ "spatial_softmax_num_keypoints": 32,
135
+ "use_separate_rgb_encoder_per_camera": true,
136
+ "hidden_size": 768,
137
+ "num_layers": 12,
138
+ "num_attention_heads": 12,
139
+ "attention_dropout": 0.1,
140
+ "patch_size": 2,
141
+ "use_adaln_zero": true,
142
+ "cross_attention_dim": 512,
143
+ "add_cross_attention": true,
144
+ "do_mask_loss_for_padding": false,
145
+ "optimizer_lr": 0.0001,
146
+ "optimizer_betas": [
147
+ 0.95,
148
+ 0.999
149
+ ],
150
+ "optimizer_eps": 1e-08,
151
+ "optimizer_weight_decay": 1e-06,
152
+ "scheduler_name": "cosine",
153
+ "scheduler_warmup_steps": 500,
154
+ "flow_matching_type": "CondOT",
155
+ "num_integration_steps": 50,
156
+ "fm_time_embed_dim": 128,
157
+ "fm_min_period": 0.004,
158
+ "fm_max_period": 4.0,
159
+ "ode_solver_method": "midpoint"
160
+ },
161
+ "output_dir": "outputs/train/2025-09-12/09-00-08_flow_matching_dit",
162
+ "job_name": "flow_matching_dit",
163
+ "resume": false,
164
+ "seed": 1000,
165
+ "num_workers": 4,
166
+ "batch_size": 8,
167
+ "steps": 200000,
168
+ "eval_freq": 20000,
169
+ "log_freq": 200,
170
+ "save_checkpoint": true,
171
+ "save_freq": 20000,
172
+ "use_policy_training_preset": true,
173
+ "optimizer": {
174
+ "type": "adam",
175
+ "lr": 0.0001,
176
+ "weight_decay": 1e-06,
177
+ "grad_clip_norm": 10.0,
178
+ "betas": [
179
+ 0.95,
180
+ 0.999
181
+ ],
182
+ "eps": 1e-08
183
+ },
184
+ "scheduler": {
185
+ "type": "diffuser",
186
+ "num_warmup_steps": 500,
187
+ "name": "cosine"
188
+ },
189
+ "eval": {
190
+ "n_episodes": 50,
191
+ "batch_size": 50,
192
+ "use_async_envs": false
193
+ },
194
+ "wandb": {
195
+ "enable": true,
196
+ "disable_artifact": false,
197
+ "project": "lerobot",
198
+ "entity": null,
199
+ "notes": null,
200
+ "run_id": "zax3nrql",
201
+ "mode": null
202
+ }
203
+ }