Hugging Face's logo

Kaz55
/
act-blue-180ep-gs500-ac60

Robotics
LeRobot
Safetensors
act
dg5f
ur5e
gelsight
ablation
Model card Files
xet
 Community
Kaz55 commited on
Commit
63a4bba
·
verified ·
1 Parent(s): 52193f8

Upload folder using huggingface_hub

Browse files
Files changed (8) hide show
  1. README.md +42 -0
  2. config.json +104 -0
  3. model.safetensors +3 -0
  4. policy_postprocessor.json +32 -0
  5. policy_postprocessor_step_0_unnormalizer_processor.safetensors +3 -0
  6. policy_preprocessor.json +80 -0
  7. policy_preprocessor_step_3_normalizer_processor.safetensors +3 -0
  8. train_config.json +237 -0
README.md ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ library_name: lerobot
3
+ pipeline_tag: robotics
4
+ tags: [act, lerobot, robotics, dg5f, ur5e, gelsight, ablation]
5
+ ---
6
+
7
+ # ACT — blue 186ep, GelSight 500x375 (native)
8
+
9
+ One point of a GelSight-resolution sweep on the DG-5F + UR5e blue-cable task.
10
+ Runs differ only in GelSight resolution, so any gap between them is attributable
11
+ to tactile resolution alone.
12
+
13
+ - **Dataset**: [Kaz55/dg5f_ur5e_blue_180ep_gs500](https://huggingface.co/datasets/Kaz55/dg5f_ur5e_blue_180ep_gs500) — 186 episodes / 186,152 frames
14
+ - **GelSight**: 500x375 (native)
15
+ - **RealSense**: 640x480 (identical across the sweep)
16
+ - **Policy**: ACT, chunk_size=60, n_action_steps=60
17
+ - **Training**: 100,000 steps (~4.3 epochs), batch 8, seed 1000
18
+
19
+ ## Inputs
20
+
21
+ `observation.state` (26) + RealSense x2 + GelSight x2
22
+
23
+ `observation.velocity` and `observation.effort` are present in the dataset but
24
+ **deliberately excluded**. Feature auto-derivation would otherwise feed them to
25
+ the policy, adding a second difference between runs and breaking the ablation.
26
+
27
+ ## Sweep (186 episodes)
28
+
29
+ | GelSight | model |
30
+ |---|---|
31
+ | 500x375 | act-blue-180ep-gs500-ac60 |
32
+ | 320x240 | act-blue-180ep-gs320-ac60 |
33
+ | 160x120 | act-blue-180ep-gs160-ac60 |
34
+ | 88x66 | act-blue-180ep-gs88-ac60 |
35
+ | none | act-blue-180ep-gs0-ac60 |
36
+
37
+ ## Caveat
38
+
39
+ On the related 180-episode `combined` sweep, final training loss was 0.118 across
40
+ *every* resolution **including no-GelSight-at-all** — training loss did not detect
41
+ the tactile input. Treat these numbers as a sanity check, not as evidence that
42
+ tactile resolution matters; that requires on-robot evaluation.
config.json ADDED
@@ -0,0 +1,104 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "act",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.images.realsense1": {
6
+ "type": "VISUAL",
7
+ "shape": [
8
+ 3,
9
+ 480,
10
+ 640
11
+ ]
12
+ },
13
+ "observation.images.realsense2": {
14
+ "type": "VISUAL",
15
+ "shape": [
16
+ 3,
17
+ 480,
18
+ 640
19
+ ]
20
+ },
21
+ "observation.images.gelsight1": {
22
+ "type": "VISUAL",
23
+ "shape": [
24
+ 3,
25
+ 375,
26
+ 500
27
+ ]
28
+ },
29
+ "observation.images.gelsight2": {
30
+ "type": "VISUAL",
31
+ "shape": [
32
+ 3,
33
+ 375,
34
+ 500
35
+ ]
36
+ },
37
+ "observation.state": {
38
+ "type": "STATE",
39
+ "shape": [
40
+ 26
41
+ ]
42
+ }
43
+ },
44
+ "output_features": {
45
+ "action": {
46
+ "type": "ACTION",
47
+ "shape": [
48
+ 26
49
+ ]
50
+ }
51
+ },
52
+ "device": "cuda",
53
+ "use_amp": false,
54
+ "use_peft": false,
55
+ "push_to_hub": false,
56
+ "repo_id": null,
57
+ "private": null,
58
+ "tags": null,
59
+ "license": null,
60
+ "pretrained_path": null,
61
+ "chunk_size": 60,
62
+ "n_action_steps": 60,
63
+ "normalization_mapping": {
64
+ "VISUAL": "MEAN_STD",
65
+ "STATE": "MEAN_STD",
66
+ "ACTION": "MEAN_STD"
67
+ },
68
+ "vision_backbone": "resnet18",
69
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
70
+ "replace_final_stride_with_dilation": false,
71
+ "pre_norm": false,
72
+ "dim_model": 512,
73
+ "n_heads": 8,
74
+ "dim_feedforward": 3200,
75
+ "feedforward_activation": "relu",
76
+ "n_encoder_layers": 4,
77
+ "n_decoder_layers": 1,
78
+ "use_vae": true,
79
+ "latent_dim": 32,
80
+ "n_vae_encoder_layers": 4,
81
+ "temporal_ensemble_coeff": null,
82
+ "dropout": 0.1,
83
+ "kl_weight": 10.0,
84
+ "optimizer_lr": 1e-05,
85
+ "optimizer_weight_decay": 0.0001,
86
+ "optimizer_lr_backbone": 1e-05,
87
+ "use_finger_tactile_encoder": false,
88
+ "thumb_tactile_key": "observation.images.gelsight2",
89
+ "index_tactile_key": "observation.images.gelsight1",
90
+ "thumb_state_indices": [
91
+ 6,
92
+ 7,
93
+ 8,
94
+ 9
95
+ ],
96
+ "index_state_indices": [
97
+ 10,
98
+ 11,
99
+ 12,
100
+ 13
101
+ ],
102
+ "tactile_keep_spatial_tokens": false,
103
+ "drop_finger_dims_from_state_token": false
104
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef80805b2f92f2ae1090ae90e3fa0ef09147df7a3924024eda2e707ba96fc58d
3
+ size 206699816
policy_postprocessor.json ADDED
@@ -0,0 +1,32 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "policy_postprocessor",
3
+ "steps": [
4
+ {
5
+ "registry_name": "unnormalizer_processor",
6
+ "config": {
7
+ "eps": 1e-08,
8
+ "features": {
9
+ "action": {
10
+ "type": "ACTION",
11
+ "shape": [
12
+ 26
13
+ ]
14
+ }
15
+ },
16
+ "norm_map": {
17
+ "VISUAL": "MEAN_STD",
18
+ "STATE": "MEAN_STD",
19
+ "ACTION": "MEAN_STD"
20
+ }
21
+ },
22
+ "state_file": "policy_postprocessor_step_0_unnormalizer_processor.safetensors"
23
+ },
24
+ {
25
+ "registry_name": "device_processor",
26
+ "config": {
27
+ "device": "cpu",
28
+ "float_dtype": null
29
+ }
30
+ }
31
+ ]
32
+ }
policy_postprocessor_step_0_unnormalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d03db6a80b5a46965b737873db838b7fb4e66658dda9ae7fbf3b090a812b7680
3
+ size 14872
policy_preprocessor.json ADDED
@@ -0,0 +1,80 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "name": "policy_preprocessor",
3
+ "steps": [
4
+ {
5
+ "registry_name": "rename_observations_processor",
6
+ "config": {
7
+ "rename_map": {}
8
+ }
9
+ },
10
+ {
11
+ "registry_name": "to_batch_processor",
12
+ "config": {}
13
+ },
14
+ {
15
+ "registry_name": "device_processor",
16
+ "config": {
17
+ "device": "cuda",
18
+ "float_dtype": null
19
+ }
20
+ },
21
+ {
22
+ "registry_name": "normalizer_processor",
23
+ "config": {
24
+ "eps": 1e-08,
25
+ "features": {
26
+ "observation.images.realsense1": {
27
+ "type": "VISUAL",
28
+ "shape": [
29
+ 3,
30
+ 480,
31
+ 640
32
+ ]
33
+ },
34
+ "observation.images.realsense2": {
35
+ "type": "VISUAL",
36
+ "shape": [
37
+ 3,
38
+ 480,
39
+ 640
40
+ ]
41
+ },
42
+ "observation.images.gelsight1": {
43
+ "type": "VISUAL",
44
+ "shape": [
45
+ 3,
46
+ 375,
47
+ 500
48
+ ]
49
+ },
50
+ "observation.images.gelsight2": {
51
+ "type": "VISUAL",
52
+ "shape": [
53
+ 3,
54
+ 375,
55
+ 500
56
+ ]
57
+ },
58
+ "observation.state": {
59
+ "type": "STATE",
60
+ "shape": [
61
+ 26
62
+ ]
63
+ },
64
+ "action": {
65
+ "type": "ACTION",
66
+ "shape": [
67
+ 26
68
+ ]
69
+ }
70
+ },
71
+ "norm_map": {
72
+ "VISUAL": "MEAN_STD",
73
+ "STATE": "MEAN_STD",
74
+ "ACTION": "MEAN_STD"
75
+ }
76
+ },
77
+ "state_file": "policy_preprocessor_step_3_normalizer_processor.safetensors"
78
+ }
79
+ ]
80
+ }
policy_preprocessor_step_3_normalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19f613519f7a03c1f3aa0905e9e363d30033147826cc28d7c3ad61eff7c466fb
3
+ size 14888
train_config.json ADDED
@@ -0,0 +1,237 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "Kaz55/dg5f_ur5e_blue_180ep_gs500",
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
+ "affine": {
62
+ "weight": 1.0,
63
+ "type": "RandomAffine",
64
+ "kwargs": {
65
+ "degrees": [
66
+ -5.0,
67
+ 5.0
68
+ ],
69
+ "translate": [
70
+ 0.05,
71
+ 0.05
72
+ ]
73
+ }
74
+ }
75
+ }
76
+ },
77
+ "revision": null,
78
+ "use_imagenet_stats": true,
79
+ "video_backend": "torchcodec",
80
+ "return_uint8": false,
81
+ "streaming": false
82
+ },
83
+ "env": null,
84
+ "policy": {
85
+ "type": "act",
86
+ "n_obs_steps": 1,
87
+ "input_features": {
88
+ "observation.images.realsense1": {
89
+ "type": "VISUAL",
90
+ "shape": [
91
+ 3,
92
+ 480,
93
+ 640
94
+ ]
95
+ },
96
+ "observation.images.realsense2": {
97
+ "type": "VISUAL",
98
+ "shape": [
99
+ 3,
100
+ 480,
101
+ 640
102
+ ]
103
+ },
104
+ "observation.images.gelsight1": {
105
+ "type": "VISUAL",
106
+ "shape": [
107
+ 3,
108
+ 375,
109
+ 500
110
+ ]
111
+ },
112
+ "observation.images.gelsight2": {
113
+ "type": "VISUAL",
114
+ "shape": [
115
+ 3,
116
+ 375,
117
+ 500
118
+ ]
119
+ },
120
+ "observation.state": {
121
+ "type": "STATE",
122
+ "shape": [
123
+ 26
124
+ ]
125
+ }
126
+ },
127
+ "output_features": {
128
+ "action": {
129
+ "type": "ACTION",
130
+ "shape": [
131
+ 26
132
+ ]
133
+ }
134
+ },
135
+ "device": "cuda",
136
+ "use_amp": false,
137
+ "use_peft": false,
138
+ "push_to_hub": false,
139
+ "repo_id": null,
140
+ "private": null,
141
+ "tags": null,
142
+ "license": null,
143
+ "pretrained_path": null,
144
+ "chunk_size": 60,
145
+ "n_action_steps": 60,
146
+ "normalization_mapping": {
147
+ "VISUAL": "MEAN_STD",
148
+ "STATE": "MEAN_STD",
149
+ "ACTION": "MEAN_STD"
150
+ },
151
+ "vision_backbone": "resnet18",
152
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
153
+ "replace_final_stride_with_dilation": false,
154
+ "pre_norm": false,
155
+ "dim_model": 512,
156
+ "n_heads": 8,
157
+ "dim_feedforward": 3200,
158
+ "feedforward_activation": "relu",
159
+ "n_encoder_layers": 4,
160
+ "n_decoder_layers": 1,
161
+ "use_vae": true,
162
+ "latent_dim": 32,
163
+ "n_vae_encoder_layers": 4,
164
+ "temporal_ensemble_coeff": null,
165
+ "dropout": 0.1,
166
+ "kl_weight": 10.0,
167
+ "optimizer_lr": 1e-05,
168
+ "optimizer_weight_decay": 0.0001,
169
+ "optimizer_lr_backbone": 1e-05,
170
+ "use_finger_tactile_encoder": false,
171
+ "thumb_tactile_key": "observation.images.gelsight2",
172
+ "index_tactile_key": "observation.images.gelsight1",
173
+ "thumb_state_indices": [
174
+ 6,
175
+ 7,
176
+ 8,
177
+ 9
178
+ ],
179
+ "index_state_indices": [
180
+ 10,
181
+ 11,
182
+ 12,
183
+ 13
184
+ ],
185
+ "tactile_keep_spatial_tokens": false,
186
+ "drop_finger_dims_from_state_token": false
187
+ },
188
+ "reward_model": null,
189
+ "output_dir": "outputs/train/act_blue_180ep_gs500_ac60",
190
+ "job_name": "act_blue_180ep_gs500_ac60",
191
+ "resume": false,
192
+ "seed": 1000,
193
+ "cudnn_deterministic": false,
194
+ "num_workers": 4,
195
+ "batch_size": 8,
196
+ "prefetch_factor": 4,
197
+ "persistent_workers": true,
198
+ "steps": 100000,
199
+ "eval_freq": 20000,
200
+ "log_freq": 200,
201
+ "tolerance_s": 0.0001,
202
+ "save_checkpoint": true,
203
+ "save_freq": 20000,
204
+ "use_policy_training_preset": true,
205
+ "optimizer": {
206
+ "type": "adamw",
207
+ "lr": 1e-05,
208
+ "weight_decay": 0.0001,
209
+ "grad_clip_norm": 10.0,
210
+ "betas": [
211
+ 0.9,
212
+ 0.999
213
+ ],
214
+ "eps": 1e-08
215
+ },
216
+ "scheduler": null,
217
+ "eval": {
218
+ "n_episodes": 50,
219
+ "batch_size": 22,
220
+ "use_async_envs": true
221
+ },
222
+ "wandb": {
223
+ "enable": false,
224
+ "disable_artifact": false,
225
+ "project": "lerobot",
226
+ "entity": null,
227
+ "notes": null,
228
+ "run_id": null,
229
+ "mode": null,
230
+ "add_tags": true
231
+ },
232
+ "peft": null,
233
+ "sample_weighting": null,
234
+ "phase_sampling": null,
235
+ "rename_map": {},
236
+ "checkpoint_path": null
237
+ }