MikelRubi commited on
Commit
b72a180
·
verified ·
1 Parent(s): 09242e7

Add pick_v3_super from IkerAzurmendi (act-pick-v3-6000, act-pick-v3-super-12000, act-pick-v3-super-18000, act-pick-v3-super-24000, act-pick-v3-super-30000)

Browse files
Files changed (45) hide show
  1. pick_v3_super/checkpoints/006000/dataset_features.json +88 -0
  2. pick_v3_super/checkpoints/006000/model_card_original.md +43 -0
  3. pick_v3_super/checkpoints/006000/pretrained_model/config.json +71 -0
  4. pick_v3_super/checkpoints/006000/pretrained_model/model.safetensors +3 -0
  5. pick_v3_super/checkpoints/006000/pretrained_model/policy_postprocessor.json +32 -0
  6. pick_v3_super/checkpoints/006000/pretrained_model/policy_postprocessor_step_0_unnormalizer_processor.safetensors +3 -0
  7. pick_v3_super/checkpoints/006000/pretrained_model/policy_preprocessor.json +64 -0
  8. pick_v3_super/checkpoints/006000/pretrained_model/policy_preprocessor_step_3_normalizer_processor.safetensors +3 -0
  9. pick_v3_super/checkpoints/006000/pretrained_model/train_config.json +600 -0
  10. pick_v3_super/checkpoints/012000/dataset_features.json +88 -0
  11. pick_v3_super/checkpoints/012000/model_card_original.md +49 -0
  12. pick_v3_super/checkpoints/012000/pretrained_model/config.json +71 -0
  13. pick_v3_super/checkpoints/012000/pretrained_model/model.safetensors +3 -0
  14. pick_v3_super/checkpoints/012000/pretrained_model/policy_postprocessor.json +32 -0
  15. pick_v3_super/checkpoints/012000/pretrained_model/policy_postprocessor_step_0_unnormalizer_processor.safetensors +3 -0
  16. pick_v3_super/checkpoints/012000/pretrained_model/policy_preprocessor.json +64 -0
  17. pick_v3_super/checkpoints/012000/pretrained_model/policy_preprocessor_step_3_normalizer_processor.safetensors +3 -0
  18. pick_v3_super/checkpoints/012000/pretrained_model/train_config.json +600 -0
  19. pick_v3_super/checkpoints/018000/dataset_features.json +88 -0
  20. pick_v3_super/checkpoints/018000/model_card_original.md +49 -0
  21. pick_v3_super/checkpoints/018000/pretrained_model/config.json +71 -0
  22. pick_v3_super/checkpoints/018000/pretrained_model/model.safetensors +3 -0
  23. pick_v3_super/checkpoints/018000/pretrained_model/policy_postprocessor.json +32 -0
  24. pick_v3_super/checkpoints/018000/pretrained_model/policy_postprocessor_step_0_unnormalizer_processor.safetensors +3 -0
  25. pick_v3_super/checkpoints/018000/pretrained_model/policy_preprocessor.json +64 -0
  26. pick_v3_super/checkpoints/018000/pretrained_model/policy_preprocessor_step_3_normalizer_processor.safetensors +3 -0
  27. pick_v3_super/checkpoints/018000/pretrained_model/train_config.json +600 -0
  28. pick_v3_super/checkpoints/024000/dataset_features.json +88 -0
  29. pick_v3_super/checkpoints/024000/model_card_original.md +49 -0
  30. pick_v3_super/checkpoints/024000/pretrained_model/config.json +71 -0
  31. pick_v3_super/checkpoints/024000/pretrained_model/model.safetensors +3 -0
  32. pick_v3_super/checkpoints/024000/pretrained_model/policy_postprocessor.json +32 -0
  33. pick_v3_super/checkpoints/024000/pretrained_model/policy_postprocessor_step_0_unnormalizer_processor.safetensors +3 -0
  34. pick_v3_super/checkpoints/024000/pretrained_model/policy_preprocessor.json +64 -0
  35. pick_v3_super/checkpoints/024000/pretrained_model/policy_preprocessor_step_3_normalizer_processor.safetensors +3 -0
  36. pick_v3_super/checkpoints/024000/pretrained_model/train_config.json +600 -0
  37. pick_v3_super/checkpoints/030000/dataset_features.json +88 -0
  38. pick_v3_super/checkpoints/030000/model_card_original.md +49 -0
  39. pick_v3_super/checkpoints/030000/pretrained_model/config.json +71 -0
  40. pick_v3_super/checkpoints/030000/pretrained_model/model.safetensors +3 -0
  41. pick_v3_super/checkpoints/030000/pretrained_model/policy_postprocessor.json +32 -0
  42. pick_v3_super/checkpoints/030000/pretrained_model/policy_postprocessor_step_0_unnormalizer_processor.safetensors +3 -0
  43. pick_v3_super/checkpoints/030000/pretrained_model/policy_preprocessor.json +64 -0
  44. pick_v3_super/checkpoints/030000/pretrained_model/policy_preprocessor_step_3_normalizer_processor.safetensors +3 -0
  45. pick_v3_super/checkpoints/030000/pretrained_model/train_config.json +600 -0
pick_v3_super/checkpoints/006000/dataset_features.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action": {
3
+ "dtype": "float32",
4
+ "names": [
5
+ "shoulder_pan.pos",
6
+ "shoulder_lift.pos",
7
+ "elbow_flex.pos",
8
+ "wrist_flex.pos",
9
+ "wrist_roll.pos",
10
+ "gripper.pos"
11
+ ],
12
+ "shape": [
13
+ 6
14
+ ]
15
+ },
16
+ "observation.images.front": {
17
+ "dtype": "video",
18
+ "shape": [
19
+ 480,
20
+ 640,
21
+ 3
22
+ ],
23
+ "names": [
24
+ "height",
25
+ "width",
26
+ "channels"
27
+ ],
28
+ "info": {
29
+ "is_depth_map": false,
30
+ "video.height": 480,
31
+ "video.width": 640,
32
+ "video.codec": "h264",
33
+ "video.pix_fmt": "yuv420p",
34
+ "video.fps": 30,
35
+ "video.channels": 3,
36
+ "has_audio": false,
37
+ "video.g": 2,
38
+ "video.crf": 30,
39
+ "video.preset": "p1",
40
+ "video.fast_decode": 0,
41
+ "video.video_backend": "pyav",
42
+ "video.extra_options": {}
43
+ }
44
+ },
45
+ "observation.images.wrist": {
46
+ "dtype": "video",
47
+ "shape": [
48
+ 480,
49
+ 640,
50
+ 3
51
+ ],
52
+ "names": [
53
+ "height",
54
+ "width",
55
+ "channels"
56
+ ],
57
+ "info": {
58
+ "is_depth_map": false,
59
+ "video.height": 480,
60
+ "video.width": 640,
61
+ "video.codec": "h264",
62
+ "video.pix_fmt": "yuv420p",
63
+ "video.fps": 30,
64
+ "video.channels": 3,
65
+ "has_audio": false,
66
+ "video.g": 2,
67
+ "video.crf": 30,
68
+ "video.preset": "p1",
69
+ "video.fast_decode": 0,
70
+ "video.video_backend": "pyav",
71
+ "video.extra_options": {}
72
+ }
73
+ },
74
+ "observation.state": {
75
+ "dtype": "float32",
76
+ "names": [
77
+ "shoulder_pan.pos",
78
+ "shoulder_lift.pos",
79
+ "elbow_flex.pos",
80
+ "wrist_flex.pos",
81
+ "wrist_roll.pos",
82
+ "gripper.pos"
83
+ ],
84
+ "shape": [
85
+ 6
86
+ ]
87
+ }
88
+ }
pick_v3_super/checkpoints/006000/model_card_original.md ADDED
@@ -0,0 +1,43 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: lerobot
4
+ pipeline_tag: robotics
5
+ tags:
6
+ - lerobot
7
+ - act
8
+ - so101
9
+ - pick
10
+ ---
11
+
12
+ # act-pick-v3-6000
13
+
14
+ Best pick policy of the third round (v3), chosen from a sweep of five training runs. ACT policy for the SO-101 "pick a piece from the crate and hold it up to the camera" task.
15
+ Inputs: front camera, wrist camera and the six joint positions; output: a chunk of 90 future joint targets at 30 fps. This is the checkpoint at 6,000 steps of a 30,000-step run (constant learning rate, so it is the same as a 6,000-step run).
16
+
17
+ ## Training data ("super" mix)
18
+
19
+ Every recording in `MikelRubi/so101-pick-cube_*` (66 sessions, including the two evening sessions of 24 Oct) that passes a quality filter:
20
+
21
+ - 5 to 15 s long, start gap between leader and follower of at most 20 degrees, at most 3 s idle at either end, gripper opens at least once;
22
+ - the known bad takes (wrong camera, people in view) removed;
23
+ - both the 30 fps recordings and the 10 fps ones (up-sampled to 30 fps), with any number of pieces in the crate.
24
+
25
+ 398 training episodes. Held out: 16 episodes, 12 from `so101-pick-cube_20261024_181939` (one piece, varied positions) and 4 from `so101-pick-cube_20261024_144537` (two pieces).
26
+
27
+ Recipe: LeRobot `train()`, batch 8, chunk 90, 30 action steps per replan, learning rate 1e-5, LeRobot's default image augmentation (colour, sharpness, small affine), seed 1000.
28
+
29
+ ## Offline check on the held-out episodes
30
+
31
+ Mean absolute joint error over the predicted chunk (lower is better) against a "stay at the current joint positions" baseline of 20.71:
32
+
33
+ | model | error | below baseline |
34
+ |---|---|---|
35
+ | **this model (super, 6,000 steps)** | **15.16** | **27%** |
36
+ | super, 12,000 / 18,000 / 24,000 / 30,000 steps | 16.08 / 16.04 / 16.36 / 16.27 | 22% / 23% / 21% / 21% |
37
+ | strict-quality run, 2,400 steps | 15.14 | 27% |
38
+ | v3 recipe on 1-2 piece episodes, 18,000 steps | 15.35 | 26% |
39
+ | v2 (`act-pick-v2-6000`) | 15.36 | 26% |
40
+
41
+ Per joint, policy / baseline: shoulder_pan 7.4 / 9.7, shoulder_lift 13.8 / 30.2, elbow_flex 16.6 / 21.0, wrist_flex 15.3 / 20.4, wrist_roll 30.8 / 28.5, gripper 7.1 / 14.9. `wrist_roll` is not better than staying still.
42
+
43
+ **This is a choice among ties.** The best checkpoints of four runs are within 0.2 points of each other, which is inside seed noise, and every run got worse after about one pass over the data. This one was picked because it has the lowest pooled error among the models trained on the new data, the most training episodes, and more training than the 2,400-step run. The offline error cannot say which model grasps best; it has not been tested on the robot.
pick_v3_super/checkpoints/006000/pretrained_model/config.json ADDED
@@ -0,0 +1,71 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "act",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 6
9
+ ]
10
+ },
11
+ "observation.images.front": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 480,
16
+ 640
17
+ ]
18
+ },
19
+ "observation.images.wrist": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 480,
24
+ 640
25
+ ]
26
+ }
27
+ },
28
+ "output_features": {
29
+ "action": {
30
+ "type": "ACTION",
31
+ "shape": [
32
+ 6
33
+ ]
34
+ }
35
+ },
36
+ "device": "cuda",
37
+ "use_amp": false,
38
+ "use_peft": false,
39
+ "push_to_hub": false,
40
+ "repo_id": null,
41
+ "private": null,
42
+ "tags": null,
43
+ "license": null,
44
+ "pretrained_path": null,
45
+ "chunk_size": 90,
46
+ "n_action_steps": 30,
47
+ "normalization_mapping": {
48
+ "VISUAL": "MEAN_STD",
49
+ "STATE": "MEAN_STD",
50
+ "ACTION": "MEAN_STD"
51
+ },
52
+ "vision_backbone": "resnet18",
53
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
54
+ "replace_final_stride_with_dilation": false,
55
+ "pre_norm": false,
56
+ "dim_model": 512,
57
+ "n_heads": 8,
58
+ "dim_feedforward": 3200,
59
+ "feedforward_activation": "relu",
60
+ "n_encoder_layers": 4,
61
+ "n_decoder_layers": 1,
62
+ "use_vae": true,
63
+ "latent_dim": 32,
64
+ "n_vae_encoder_layers": 4,
65
+ "temporal_ensemble_coeff": null,
66
+ "dropout": 0.1,
67
+ "kl_weight": 10.0,
68
+ "optimizer_lr": 1e-05,
69
+ "optimizer_weight_decay": 0.0001,
70
+ "optimizer_lr_backbone": 1e-05
71
+ }
pick_v3_super/checkpoints/006000/pretrained_model/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b57d1555a52932f3260cf9a92238066b20cea3b3b0e0039468f4c3194cadd6ca
3
+ size 206658768
pick_v3_super/checkpoints/006000/pretrained_model/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
+ 6
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
+ }
pick_v3_super/checkpoints/006000/pretrained_model/policy_postprocessor_step_0_unnormalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931ed929c95ef00f6e334882511164d4c9d89f16bfb31eeb212ddecf99897ed7
3
+ size 7552
pick_v3_super/checkpoints/006000/pretrained_model/policy_preprocessor.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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.state": {
27
+ "type": "STATE",
28
+ "shape": [
29
+ 6
30
+ ]
31
+ },
32
+ "observation.images.front": {
33
+ "type": "VISUAL",
34
+ "shape": [
35
+ 3,
36
+ 480,
37
+ 640
38
+ ]
39
+ },
40
+ "observation.images.wrist": {
41
+ "type": "VISUAL",
42
+ "shape": [
43
+ 3,
44
+ 480,
45
+ 640
46
+ ]
47
+ },
48
+ "action": {
49
+ "type": "ACTION",
50
+ "shape": [
51
+ 6
52
+ ]
53
+ }
54
+ },
55
+ "norm_map": {
56
+ "VISUAL": "MEAN_STD",
57
+ "STATE": "MEAN_STD",
58
+ "ACTION": "MEAN_STD"
59
+ }
60
+ },
61
+ "state_file": "policy_preprocessor_step_3_normalizer_processor.safetensors"
62
+ }
63
+ ]
64
+ }
pick_v3_super/checkpoints/006000/pretrained_model/policy_preprocessor_step_3_normalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931ed929c95ef00f6e334882511164d4c9d89f16bfb31eeb212ddecf99897ed7
3
+ size 7552
pick_v3_super/checkpoints/006000/pretrained_model/train_config.json ADDED
@@ -0,0 +1,600 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "local/so101_act",
4
+ "root": "C:\\Users\\Usuario\\Documents\\Workspace\\LeRobot_\\demos\\06_imitation_learning\\data\\real_so101\\act_pick_super",
5
+ "episodes": [
6
+ 4,
7
+ 5,
8
+ 6,
9
+ 7,
10
+ 8,
11
+ 9,
12
+ 10,
13
+ 12,
14
+ 15,
15
+ 17,
16
+ 18,
17
+ 19,
18
+ 20,
19
+ 21,
20
+ 23,
21
+ 24,
22
+ 25,
23
+ 27,
24
+ 28,
25
+ 29,
26
+ 30,
27
+ 31,
28
+ 32,
29
+ 33,
30
+ 34,
31
+ 35,
32
+ 36,
33
+ 37,
34
+ 38,
35
+ 39,
36
+ 41,
37
+ 42,
38
+ 43,
39
+ 44,
40
+ 45,
41
+ 46,
42
+ 47,
43
+ 49,
44
+ 50,
45
+ 51,
46
+ 52,
47
+ 53,
48
+ 54,
49
+ 55,
50
+ 56,
51
+ 57,
52
+ 58,
53
+ 59,
54
+ 60,
55
+ 63,
56
+ 64,
57
+ 65,
58
+ 66,
59
+ 67,
60
+ 68,
61
+ 69,
62
+ 70,
63
+ 71,
64
+ 72,
65
+ 73,
66
+ 74,
67
+ 76,
68
+ 77,
69
+ 78,
70
+ 79,
71
+ 84,
72
+ 86,
73
+ 87,
74
+ 88,
75
+ 89,
76
+ 90,
77
+ 91,
78
+ 92,
79
+ 93,
80
+ 94,
81
+ 96,
82
+ 97,
83
+ 98,
84
+ 100,
85
+ 101,
86
+ 102,
87
+ 103,
88
+ 104,
89
+ 105,
90
+ 106,
91
+ 107,
92
+ 109,
93
+ 110,
94
+ 111,
95
+ 112,
96
+ 113,
97
+ 114,
98
+ 115,
99
+ 116,
100
+ 117,
101
+ 118,
102
+ 119,
103
+ 120,
104
+ 121,
105
+ 122,
106
+ 124,
107
+ 125,
108
+ 126,
109
+ 127,
110
+ 128,
111
+ 129,
112
+ 130,
113
+ 131,
114
+ 132,
115
+ 133,
116
+ 134,
117
+ 135,
118
+ 136,
119
+ 137,
120
+ 138,
121
+ 139,
122
+ 140,
123
+ 141,
124
+ 142,
125
+ 143,
126
+ 144,
127
+ 145,
128
+ 146,
129
+ 147,
130
+ 148,
131
+ 149,
132
+ 150,
133
+ 151,
134
+ 152,
135
+ 153,
136
+ 154,
137
+ 155,
138
+ 156,
139
+ 157,
140
+ 158,
141
+ 159,
142
+ 160,
143
+ 161,
144
+ 162,
145
+ 163,
146
+ 164,
147
+ 165,
148
+ 166,
149
+ 167,
150
+ 168,
151
+ 169,
152
+ 170,
153
+ 171,
154
+ 172,
155
+ 173,
156
+ 174,
157
+ 175,
158
+ 176,
159
+ 177,
160
+ 178,
161
+ 179,
162
+ 180,
163
+ 181,
164
+ 182,
165
+ 183,
166
+ 184,
167
+ 185,
168
+ 186,
169
+ 187,
170
+ 188,
171
+ 189,
172
+ 190,
173
+ 191,
174
+ 192,
175
+ 193,
176
+ 194,
177
+ 195,
178
+ 196,
179
+ 197,
180
+ 198,
181
+ 199,
182
+ 200,
183
+ 201,
184
+ 202,
185
+ 203,
186
+ 204,
187
+ 205,
188
+ 206,
189
+ 207,
190
+ 208,
191
+ 209,
192
+ 210,
193
+ 211,
194
+ 212,
195
+ 213,
196
+ 214,
197
+ 215,
198
+ 216,
199
+ 217,
200
+ 218,
201
+ 219,
202
+ 220,
203
+ 221,
204
+ 222,
205
+ 223,
206
+ 224,
207
+ 225,
208
+ 226,
209
+ 227,
210
+ 228,
211
+ 229,
212
+ 230,
213
+ 231,
214
+ 232,
215
+ 233,
216
+ 235,
217
+ 236,
218
+ 237,
219
+ 238,
220
+ 239,
221
+ 240,
222
+ 241,
223
+ 242,
224
+ 243,
225
+ 244,
226
+ 245,
227
+ 246,
228
+ 247,
229
+ 248,
230
+ 249,
231
+ 250,
232
+ 251,
233
+ 252,
234
+ 253,
235
+ 255,
236
+ 256,
237
+ 257,
238
+ 258,
239
+ 259,
240
+ 260,
241
+ 261,
242
+ 262,
243
+ 263,
244
+ 264,
245
+ 265,
246
+ 266,
247
+ 267,
248
+ 268,
249
+ 269,
250
+ 270,
251
+ 271,
252
+ 272,
253
+ 273,
254
+ 274,
255
+ 275,
256
+ 276,
257
+ 277,
258
+ 278,
259
+ 279,
260
+ 280,
261
+ 281,
262
+ 282,
263
+ 283,
264
+ 284,
265
+ 285,
266
+ 286,
267
+ 287,
268
+ 288,
269
+ 289,
270
+ 290,
271
+ 291,
272
+ 292,
273
+ 293,
274
+ 294,
275
+ 295,
276
+ 296,
277
+ 297,
278
+ 298,
279
+ 299,
280
+ 300,
281
+ 301,
282
+ 302,
283
+ 303,
284
+ 304,
285
+ 305,
286
+ 306,
287
+ 307,
288
+ 308,
289
+ 309,
290
+ 310,
291
+ 311,
292
+ 312,
293
+ 313,
294
+ 314,
295
+ 315,
296
+ 316,
297
+ 317,
298
+ 318,
299
+ 319,
300
+ 320,
301
+ 325,
302
+ 326,
303
+ 327,
304
+ 329,
305
+ 330,
306
+ 331,
307
+ 332,
308
+ 333,
309
+ 334,
310
+ 335,
311
+ 336,
312
+ 337,
313
+ 338,
314
+ 339,
315
+ 340,
316
+ 341,
317
+ 342,
318
+ 343,
319
+ 344,
320
+ 345,
321
+ 346,
322
+ 347,
323
+ 348,
324
+ 349,
325
+ 351,
326
+ 352,
327
+ 353,
328
+ 354,
329
+ 355,
330
+ 356,
331
+ 357,
332
+ 358,
333
+ 359,
334
+ 360,
335
+ 361,
336
+ 364,
337
+ 365,
338
+ 366,
339
+ 367,
340
+ 368,
341
+ 369,
342
+ 370,
343
+ 371,
344
+ 372,
345
+ 373,
346
+ 374,
347
+ 375,
348
+ 376,
349
+ 377,
350
+ 378,
351
+ 391,
352
+ 392,
353
+ 393,
354
+ 394,
355
+ 395,
356
+ 396,
357
+ 397,
358
+ 398,
359
+ 399,
360
+ 400,
361
+ 401,
362
+ 402,
363
+ 403,
364
+ 404,
365
+ 405,
366
+ 406,
367
+ 407,
368
+ 408,
369
+ 409,
370
+ 410,
371
+ 411,
372
+ 412,
373
+ 413,
374
+ 414,
375
+ 415,
376
+ 416,
377
+ 417,
378
+ 419,
379
+ 420,
380
+ 421,
381
+ 422,
382
+ 423,
383
+ 424,
384
+ 425,
385
+ 426,
386
+ 427,
387
+ 428,
388
+ 429,
389
+ 430,
390
+ 431,
391
+ 432,
392
+ 433,
393
+ 434,
394
+ 435,
395
+ 436,
396
+ 438,
397
+ 439,
398
+ 440,
399
+ 441,
400
+ 442,
401
+ 443,
402
+ 444,
403
+ 445
404
+ ],
405
+ "image_transforms": {
406
+ "enable": true,
407
+ "max_num_transforms": 3,
408
+ "random_order": false,
409
+ "tfs": {
410
+ "brightness": {
411
+ "weight": 1.0,
412
+ "type": "ColorJitter",
413
+ "kwargs": {
414
+ "brightness": [
415
+ 0.8,
416
+ 1.2
417
+ ]
418
+ }
419
+ },
420
+ "contrast": {
421
+ "weight": 1.0,
422
+ "type": "ColorJitter",
423
+ "kwargs": {
424
+ "contrast": [
425
+ 0.8,
426
+ 1.2
427
+ ]
428
+ }
429
+ },
430
+ "saturation": {
431
+ "weight": 1.0,
432
+ "type": "ColorJitter",
433
+ "kwargs": {
434
+ "saturation": [
435
+ 0.5,
436
+ 1.5
437
+ ]
438
+ }
439
+ },
440
+ "hue": {
441
+ "weight": 1.0,
442
+ "type": "ColorJitter",
443
+ "kwargs": {
444
+ "hue": [
445
+ -0.05,
446
+ 0.05
447
+ ]
448
+ }
449
+ },
450
+ "sharpness": {
451
+ "weight": 1.0,
452
+ "type": "SharpnessJitter",
453
+ "kwargs": {
454
+ "sharpness": [
455
+ 0.5,
456
+ 1.5
457
+ ]
458
+ }
459
+ },
460
+ "affine": {
461
+ "weight": 1.0,
462
+ "type": "RandomAffine",
463
+ "kwargs": {
464
+ "degrees": [
465
+ -5.0,
466
+ 5.0
467
+ ],
468
+ "translate": [
469
+ 0.05,
470
+ 0.05
471
+ ]
472
+ }
473
+ }
474
+ }
475
+ },
476
+ "revision": null,
477
+ "use_imagenet_stats": true,
478
+ "video_backend": "pyav",
479
+ "streaming": false
480
+ },
481
+ "env": null,
482
+ "policy": {
483
+ "type": "act",
484
+ "n_obs_steps": 1,
485
+ "input_features": {
486
+ "observation.state": {
487
+ "type": "STATE",
488
+ "shape": [
489
+ 6
490
+ ]
491
+ },
492
+ "observation.images.front": {
493
+ "type": "VISUAL",
494
+ "shape": [
495
+ 3,
496
+ 480,
497
+ 640
498
+ ]
499
+ },
500
+ "observation.images.wrist": {
501
+ "type": "VISUAL",
502
+ "shape": [
503
+ 3,
504
+ 480,
505
+ 640
506
+ ]
507
+ }
508
+ },
509
+ "output_features": {
510
+ "action": {
511
+ "type": "ACTION",
512
+ "shape": [
513
+ 6
514
+ ]
515
+ }
516
+ },
517
+ "device": "cuda",
518
+ "use_amp": false,
519
+ "use_peft": false,
520
+ "push_to_hub": false,
521
+ "repo_id": null,
522
+ "private": null,
523
+ "tags": null,
524
+ "license": null,
525
+ "pretrained_path": null,
526
+ "chunk_size": 90,
527
+ "n_action_steps": 30,
528
+ "normalization_mapping": {
529
+ "VISUAL": "MEAN_STD",
530
+ "STATE": "MEAN_STD",
531
+ "ACTION": "MEAN_STD"
532
+ },
533
+ "vision_backbone": "resnet18",
534
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
535
+ "replace_final_stride_with_dilation": false,
536
+ "pre_norm": false,
537
+ "dim_model": 512,
538
+ "n_heads": 8,
539
+ "dim_feedforward": 3200,
540
+ "feedforward_activation": "relu",
541
+ "n_encoder_layers": 4,
542
+ "n_decoder_layers": 1,
543
+ "use_vae": true,
544
+ "latent_dim": 32,
545
+ "n_vae_encoder_layers": 4,
546
+ "temporal_ensemble_coeff": null,
547
+ "dropout": 0.1,
548
+ "kl_weight": 10.0,
549
+ "optimizer_lr": 1e-05,
550
+ "optimizer_weight_decay": 0.0001,
551
+ "optimizer_lr_backbone": 1e-05
552
+ },
553
+ "output_dir": "C:\\Users\\Usuario\\Documents\\Workspace\\LeRobot_\\demos\\06_imitation_learning\\outputs\\sweep_super",
554
+ "job_name": "act_so101_sweep_super",
555
+ "resume": false,
556
+ "seed": 1000,
557
+ "num_workers": 4,
558
+ "batch_size": 8,
559
+ "steps": 30000,
560
+ "eval_freq": 20000,
561
+ "log_freq": 100,
562
+ "tolerance_s": 0.0001,
563
+ "save_checkpoint": true,
564
+ "save_freq": 6000,
565
+ "use_policy_training_preset": true,
566
+ "optimizer": {
567
+ "type": "adamw",
568
+ "lr": 1e-05,
569
+ "weight_decay": 0.0001,
570
+ "grad_clip_norm": 10.0,
571
+ "betas": [
572
+ 0.9,
573
+ 0.999
574
+ ],
575
+ "eps": 1e-08
576
+ },
577
+ "scheduler": null,
578
+ "eval": {
579
+ "n_episodes": 50,
580
+ "batch_size": 50,
581
+ "use_async_envs": false
582
+ },
583
+ "wandb": {
584
+ "enable": false,
585
+ "disable_artifact": false,
586
+ "project": "lerobot",
587
+ "entity": null,
588
+ "notes": null,
589
+ "run_id": null,
590
+ "mode": null
591
+ },
592
+ "peft": null,
593
+ "use_rabc": false,
594
+ "rabc_progress_path": null,
595
+ "rabc_kappa": 0.01,
596
+ "rabc_epsilon": 1e-06,
597
+ "rabc_head_mode": "sparse",
598
+ "rename_map": {},
599
+ "checkpoint_path": null
600
+ }
pick_v3_super/checkpoints/012000/dataset_features.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action": {
3
+ "dtype": "float32",
4
+ "names": [
5
+ "shoulder_pan.pos",
6
+ "shoulder_lift.pos",
7
+ "elbow_flex.pos",
8
+ "wrist_flex.pos",
9
+ "wrist_roll.pos",
10
+ "gripper.pos"
11
+ ],
12
+ "shape": [
13
+ 6
14
+ ]
15
+ },
16
+ "observation.images.front": {
17
+ "dtype": "video",
18
+ "shape": [
19
+ 480,
20
+ 640,
21
+ 3
22
+ ],
23
+ "names": [
24
+ "height",
25
+ "width",
26
+ "channels"
27
+ ],
28
+ "info": {
29
+ "is_depth_map": false,
30
+ "video.height": 480,
31
+ "video.width": 640,
32
+ "video.codec": "h264",
33
+ "video.pix_fmt": "yuv420p",
34
+ "video.fps": 30,
35
+ "video.channels": 3,
36
+ "has_audio": false,
37
+ "video.g": 2,
38
+ "video.crf": 30,
39
+ "video.preset": "p1",
40
+ "video.fast_decode": 0,
41
+ "video.video_backend": "pyav",
42
+ "video.extra_options": {}
43
+ }
44
+ },
45
+ "observation.images.wrist": {
46
+ "dtype": "video",
47
+ "shape": [
48
+ 480,
49
+ 640,
50
+ 3
51
+ ],
52
+ "names": [
53
+ "height",
54
+ "width",
55
+ "channels"
56
+ ],
57
+ "info": {
58
+ "is_depth_map": false,
59
+ "video.height": 480,
60
+ "video.width": 640,
61
+ "video.codec": "h264",
62
+ "video.pix_fmt": "yuv420p",
63
+ "video.fps": 30,
64
+ "video.channels": 3,
65
+ "has_audio": false,
66
+ "video.g": 2,
67
+ "video.crf": 30,
68
+ "video.preset": "p1",
69
+ "video.fast_decode": 0,
70
+ "video.video_backend": "pyav",
71
+ "video.extra_options": {}
72
+ }
73
+ },
74
+ "observation.state": {
75
+ "dtype": "float32",
76
+ "names": [
77
+ "shoulder_pan.pos",
78
+ "shoulder_lift.pos",
79
+ "elbow_flex.pos",
80
+ "wrist_flex.pos",
81
+ "wrist_roll.pos",
82
+ "gripper.pos"
83
+ ],
84
+ "shape": [
85
+ 6
86
+ ]
87
+ }
88
+ }
pick_v3_super/checkpoints/012000/model_card_original.md ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: lerobot
4
+ pipeline_tag: robotics
5
+ tags:
6
+ - lerobot
7
+ - act
8
+ - so101
9
+ - pick
10
+ ---
11
+
12
+ # act-pick-v3-super-12000
13
+
14
+ Pick policy trained on the "super" mix of the third round (v3): every recording that passes the quality filter. Checkpoint at 12,000 steps. This is not the best checkpoint of its run (the best one is at 6,000 steps); it is published for completeness.
15
+
16
+ ACT policy for the SO-101 "pick a piece from the crate and hold it up to the camera" task. Inputs: front camera, wrist camera and the six joint positions; output: a chunk of 90 future joint targets at 30 fps.
17
+
18
+ ## Training data
19
+
20
+ 398 training episodes: all 66 pick-cube sessions, 30 fps and 10 fps (up-sampled to 30 fps), any number of pieces in the crate; 5 to 15 s long, start gap at most 20 degrees, at most 3 s idle at either end, gripper opens at least once, known bad takes removed.
21
+
22
+ Recipe: LeRobot `train()`, batch 8, chunk 90, 30 action steps per replan, default image augmentation, seed 1000.
23
+
24
+ ## Offline check on the held-out episodes
25
+
26
+ Held out: 16 episodes, 12 from `so101-pick-cube_20261024_181939` (one piece) and 4 from `so101-pick-cube_20261024_144537` (two pieces). Mean absolute joint error over the predicted chunk (lower is better) against a "stay at the current joint positions" baseline of 20.71:
27
+
28
+ | checkpoint of this run | error | below baseline |
29
+ |---|---|---|
30
+ | 6,000 steps | 15.16 | 27% |
31
+ | **12,000 steps (this one)** | **16.08** | 22% |
32
+ | 18,000 steps | 16.04 | 23% |
33
+ | 24,000 steps | 16.36 | 21% |
34
+ | 30,000 steps | 16.27 | 21% |
35
+
36
+ All runs of this round, best checkpoint of each:
37
+
38
+ | run | training episodes | best checkpoint | error | below stay-still | repo |
39
+ |---|---|---|---|---|---|
40
+ | super (all quality-passing data) | 398 | 6k | 15.16 | 27% | act-pick-v3-6000 |
41
+ | strict (6-12 s, tight quality) | 205 | 2.4k | 15.14 | 27% | act-pick-v3-strict-2400 |
42
+ | v3 recipe on 1-2 piece episodes | 331 | 18k | 15.35 | 26% | act-pick-v3-1-2-pieces-18000 |
43
+ | v2 (earlier) | 278 | 6k | 15.36 | 26% | act-pick-v2-6000 |
44
+ | chunk 60 (own baseline 16.23) | 331 | 4.8k | 11.94 | 26% | act-pick-v3-chunk60-4800 |
45
+ | one piece only | 292 | 9.6k | 15.47 | 25% | act-pick-v3-1piece-9600 |
46
+ | recent sessions only | 100 | 12k | 16.87 | 19% | act-pick-v3-recent-12000 |
47
+ | first run | 165 | 6k | 16.87 | 19% | act-pick-* |
48
+
49
+ The best checkpoints of the main runs are within noise of each other, and every run got worse after about one pass over the data. The offline error cannot say which model grasps best; none of them has been tested on the robot.
pick_v3_super/checkpoints/012000/pretrained_model/config.json ADDED
@@ -0,0 +1,71 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "act",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 6
9
+ ]
10
+ },
11
+ "observation.images.front": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 480,
16
+ 640
17
+ ]
18
+ },
19
+ "observation.images.wrist": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 480,
24
+ 640
25
+ ]
26
+ }
27
+ },
28
+ "output_features": {
29
+ "action": {
30
+ "type": "ACTION",
31
+ "shape": [
32
+ 6
33
+ ]
34
+ }
35
+ },
36
+ "device": "cuda",
37
+ "use_amp": false,
38
+ "use_peft": false,
39
+ "push_to_hub": false,
40
+ "repo_id": null,
41
+ "private": null,
42
+ "tags": null,
43
+ "license": null,
44
+ "pretrained_path": null,
45
+ "chunk_size": 90,
46
+ "n_action_steps": 30,
47
+ "normalization_mapping": {
48
+ "VISUAL": "MEAN_STD",
49
+ "STATE": "MEAN_STD",
50
+ "ACTION": "MEAN_STD"
51
+ },
52
+ "vision_backbone": "resnet18",
53
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
54
+ "replace_final_stride_with_dilation": false,
55
+ "pre_norm": false,
56
+ "dim_model": 512,
57
+ "n_heads": 8,
58
+ "dim_feedforward": 3200,
59
+ "feedforward_activation": "relu",
60
+ "n_encoder_layers": 4,
61
+ "n_decoder_layers": 1,
62
+ "use_vae": true,
63
+ "latent_dim": 32,
64
+ "n_vae_encoder_layers": 4,
65
+ "temporal_ensemble_coeff": null,
66
+ "dropout": 0.1,
67
+ "kl_weight": 10.0,
68
+ "optimizer_lr": 1e-05,
69
+ "optimizer_weight_decay": 0.0001,
70
+ "optimizer_lr_backbone": 1e-05
71
+ }
pick_v3_super/checkpoints/012000/pretrained_model/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3cf9f7cb04f8f2325beb4cffcbfd142011e195449ae9eb2582a5aaa71d431491
3
+ size 206658768
pick_v3_super/checkpoints/012000/pretrained_model/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
+ 6
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
+ }
pick_v3_super/checkpoints/012000/pretrained_model/policy_postprocessor_step_0_unnormalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931ed929c95ef00f6e334882511164d4c9d89f16bfb31eeb212ddecf99897ed7
3
+ size 7552
pick_v3_super/checkpoints/012000/pretrained_model/policy_preprocessor.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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.state": {
27
+ "type": "STATE",
28
+ "shape": [
29
+ 6
30
+ ]
31
+ },
32
+ "observation.images.front": {
33
+ "type": "VISUAL",
34
+ "shape": [
35
+ 3,
36
+ 480,
37
+ 640
38
+ ]
39
+ },
40
+ "observation.images.wrist": {
41
+ "type": "VISUAL",
42
+ "shape": [
43
+ 3,
44
+ 480,
45
+ 640
46
+ ]
47
+ },
48
+ "action": {
49
+ "type": "ACTION",
50
+ "shape": [
51
+ 6
52
+ ]
53
+ }
54
+ },
55
+ "norm_map": {
56
+ "VISUAL": "MEAN_STD",
57
+ "STATE": "MEAN_STD",
58
+ "ACTION": "MEAN_STD"
59
+ }
60
+ },
61
+ "state_file": "policy_preprocessor_step_3_normalizer_processor.safetensors"
62
+ }
63
+ ]
64
+ }
pick_v3_super/checkpoints/012000/pretrained_model/policy_preprocessor_step_3_normalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931ed929c95ef00f6e334882511164d4c9d89f16bfb31eeb212ddecf99897ed7
3
+ size 7552
pick_v3_super/checkpoints/012000/pretrained_model/train_config.json ADDED
@@ -0,0 +1,600 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "local/so101_act",
4
+ "root": "C:\\Users\\Usuario\\Documents\\Workspace\\LeRobot_\\demos\\06_imitation_learning\\data\\real_so101\\act_pick_super",
5
+ "episodes": [
6
+ 4,
7
+ 5,
8
+ 6,
9
+ 7,
10
+ 8,
11
+ 9,
12
+ 10,
13
+ 12,
14
+ 15,
15
+ 17,
16
+ 18,
17
+ 19,
18
+ 20,
19
+ 21,
20
+ 23,
21
+ 24,
22
+ 25,
23
+ 27,
24
+ 28,
25
+ 29,
26
+ 30,
27
+ 31,
28
+ 32,
29
+ 33,
30
+ 34,
31
+ 35,
32
+ 36,
33
+ 37,
34
+ 38,
35
+ 39,
36
+ 41,
37
+ 42,
38
+ 43,
39
+ 44,
40
+ 45,
41
+ 46,
42
+ 47,
43
+ 49,
44
+ 50,
45
+ 51,
46
+ 52,
47
+ 53,
48
+ 54,
49
+ 55,
50
+ 56,
51
+ 57,
52
+ 58,
53
+ 59,
54
+ 60,
55
+ 63,
56
+ 64,
57
+ 65,
58
+ 66,
59
+ 67,
60
+ 68,
61
+ 69,
62
+ 70,
63
+ 71,
64
+ 72,
65
+ 73,
66
+ 74,
67
+ 76,
68
+ 77,
69
+ 78,
70
+ 79,
71
+ 84,
72
+ 86,
73
+ 87,
74
+ 88,
75
+ 89,
76
+ 90,
77
+ 91,
78
+ 92,
79
+ 93,
80
+ 94,
81
+ 96,
82
+ 97,
83
+ 98,
84
+ 100,
85
+ 101,
86
+ 102,
87
+ 103,
88
+ 104,
89
+ 105,
90
+ 106,
91
+ 107,
92
+ 109,
93
+ 110,
94
+ 111,
95
+ 112,
96
+ 113,
97
+ 114,
98
+ 115,
99
+ 116,
100
+ 117,
101
+ 118,
102
+ 119,
103
+ 120,
104
+ 121,
105
+ 122,
106
+ 124,
107
+ 125,
108
+ 126,
109
+ 127,
110
+ 128,
111
+ 129,
112
+ 130,
113
+ 131,
114
+ 132,
115
+ 133,
116
+ 134,
117
+ 135,
118
+ 136,
119
+ 137,
120
+ 138,
121
+ 139,
122
+ 140,
123
+ 141,
124
+ 142,
125
+ 143,
126
+ 144,
127
+ 145,
128
+ 146,
129
+ 147,
130
+ 148,
131
+ 149,
132
+ 150,
133
+ 151,
134
+ 152,
135
+ 153,
136
+ 154,
137
+ 155,
138
+ 156,
139
+ 157,
140
+ 158,
141
+ 159,
142
+ 160,
143
+ 161,
144
+ 162,
145
+ 163,
146
+ 164,
147
+ 165,
148
+ 166,
149
+ 167,
150
+ 168,
151
+ 169,
152
+ 170,
153
+ 171,
154
+ 172,
155
+ 173,
156
+ 174,
157
+ 175,
158
+ 176,
159
+ 177,
160
+ 178,
161
+ 179,
162
+ 180,
163
+ 181,
164
+ 182,
165
+ 183,
166
+ 184,
167
+ 185,
168
+ 186,
169
+ 187,
170
+ 188,
171
+ 189,
172
+ 190,
173
+ 191,
174
+ 192,
175
+ 193,
176
+ 194,
177
+ 195,
178
+ 196,
179
+ 197,
180
+ 198,
181
+ 199,
182
+ 200,
183
+ 201,
184
+ 202,
185
+ 203,
186
+ 204,
187
+ 205,
188
+ 206,
189
+ 207,
190
+ 208,
191
+ 209,
192
+ 210,
193
+ 211,
194
+ 212,
195
+ 213,
196
+ 214,
197
+ 215,
198
+ 216,
199
+ 217,
200
+ 218,
201
+ 219,
202
+ 220,
203
+ 221,
204
+ 222,
205
+ 223,
206
+ 224,
207
+ 225,
208
+ 226,
209
+ 227,
210
+ 228,
211
+ 229,
212
+ 230,
213
+ 231,
214
+ 232,
215
+ 233,
216
+ 235,
217
+ 236,
218
+ 237,
219
+ 238,
220
+ 239,
221
+ 240,
222
+ 241,
223
+ 242,
224
+ 243,
225
+ 244,
226
+ 245,
227
+ 246,
228
+ 247,
229
+ 248,
230
+ 249,
231
+ 250,
232
+ 251,
233
+ 252,
234
+ 253,
235
+ 255,
236
+ 256,
237
+ 257,
238
+ 258,
239
+ 259,
240
+ 260,
241
+ 261,
242
+ 262,
243
+ 263,
244
+ 264,
245
+ 265,
246
+ 266,
247
+ 267,
248
+ 268,
249
+ 269,
250
+ 270,
251
+ 271,
252
+ 272,
253
+ 273,
254
+ 274,
255
+ 275,
256
+ 276,
257
+ 277,
258
+ 278,
259
+ 279,
260
+ 280,
261
+ 281,
262
+ 282,
263
+ 283,
264
+ 284,
265
+ 285,
266
+ 286,
267
+ 287,
268
+ 288,
269
+ 289,
270
+ 290,
271
+ 291,
272
+ 292,
273
+ 293,
274
+ 294,
275
+ 295,
276
+ 296,
277
+ 297,
278
+ 298,
279
+ 299,
280
+ 300,
281
+ 301,
282
+ 302,
283
+ 303,
284
+ 304,
285
+ 305,
286
+ 306,
287
+ 307,
288
+ 308,
289
+ 309,
290
+ 310,
291
+ 311,
292
+ 312,
293
+ 313,
294
+ 314,
295
+ 315,
296
+ 316,
297
+ 317,
298
+ 318,
299
+ 319,
300
+ 320,
301
+ 325,
302
+ 326,
303
+ 327,
304
+ 329,
305
+ 330,
306
+ 331,
307
+ 332,
308
+ 333,
309
+ 334,
310
+ 335,
311
+ 336,
312
+ 337,
313
+ 338,
314
+ 339,
315
+ 340,
316
+ 341,
317
+ 342,
318
+ 343,
319
+ 344,
320
+ 345,
321
+ 346,
322
+ 347,
323
+ 348,
324
+ 349,
325
+ 351,
326
+ 352,
327
+ 353,
328
+ 354,
329
+ 355,
330
+ 356,
331
+ 357,
332
+ 358,
333
+ 359,
334
+ 360,
335
+ 361,
336
+ 364,
337
+ 365,
338
+ 366,
339
+ 367,
340
+ 368,
341
+ 369,
342
+ 370,
343
+ 371,
344
+ 372,
345
+ 373,
346
+ 374,
347
+ 375,
348
+ 376,
349
+ 377,
350
+ 378,
351
+ 391,
352
+ 392,
353
+ 393,
354
+ 394,
355
+ 395,
356
+ 396,
357
+ 397,
358
+ 398,
359
+ 399,
360
+ 400,
361
+ 401,
362
+ 402,
363
+ 403,
364
+ 404,
365
+ 405,
366
+ 406,
367
+ 407,
368
+ 408,
369
+ 409,
370
+ 410,
371
+ 411,
372
+ 412,
373
+ 413,
374
+ 414,
375
+ 415,
376
+ 416,
377
+ 417,
378
+ 419,
379
+ 420,
380
+ 421,
381
+ 422,
382
+ 423,
383
+ 424,
384
+ 425,
385
+ 426,
386
+ 427,
387
+ 428,
388
+ 429,
389
+ 430,
390
+ 431,
391
+ 432,
392
+ 433,
393
+ 434,
394
+ 435,
395
+ 436,
396
+ 438,
397
+ 439,
398
+ 440,
399
+ 441,
400
+ 442,
401
+ 443,
402
+ 444,
403
+ 445
404
+ ],
405
+ "image_transforms": {
406
+ "enable": true,
407
+ "max_num_transforms": 3,
408
+ "random_order": false,
409
+ "tfs": {
410
+ "brightness": {
411
+ "weight": 1.0,
412
+ "type": "ColorJitter",
413
+ "kwargs": {
414
+ "brightness": [
415
+ 0.8,
416
+ 1.2
417
+ ]
418
+ }
419
+ },
420
+ "contrast": {
421
+ "weight": 1.0,
422
+ "type": "ColorJitter",
423
+ "kwargs": {
424
+ "contrast": [
425
+ 0.8,
426
+ 1.2
427
+ ]
428
+ }
429
+ },
430
+ "saturation": {
431
+ "weight": 1.0,
432
+ "type": "ColorJitter",
433
+ "kwargs": {
434
+ "saturation": [
435
+ 0.5,
436
+ 1.5
437
+ ]
438
+ }
439
+ },
440
+ "hue": {
441
+ "weight": 1.0,
442
+ "type": "ColorJitter",
443
+ "kwargs": {
444
+ "hue": [
445
+ -0.05,
446
+ 0.05
447
+ ]
448
+ }
449
+ },
450
+ "sharpness": {
451
+ "weight": 1.0,
452
+ "type": "SharpnessJitter",
453
+ "kwargs": {
454
+ "sharpness": [
455
+ 0.5,
456
+ 1.5
457
+ ]
458
+ }
459
+ },
460
+ "affine": {
461
+ "weight": 1.0,
462
+ "type": "RandomAffine",
463
+ "kwargs": {
464
+ "degrees": [
465
+ -5.0,
466
+ 5.0
467
+ ],
468
+ "translate": [
469
+ 0.05,
470
+ 0.05
471
+ ]
472
+ }
473
+ }
474
+ }
475
+ },
476
+ "revision": null,
477
+ "use_imagenet_stats": true,
478
+ "video_backend": "pyav",
479
+ "streaming": false
480
+ },
481
+ "env": null,
482
+ "policy": {
483
+ "type": "act",
484
+ "n_obs_steps": 1,
485
+ "input_features": {
486
+ "observation.state": {
487
+ "type": "STATE",
488
+ "shape": [
489
+ 6
490
+ ]
491
+ },
492
+ "observation.images.front": {
493
+ "type": "VISUAL",
494
+ "shape": [
495
+ 3,
496
+ 480,
497
+ 640
498
+ ]
499
+ },
500
+ "observation.images.wrist": {
501
+ "type": "VISUAL",
502
+ "shape": [
503
+ 3,
504
+ 480,
505
+ 640
506
+ ]
507
+ }
508
+ },
509
+ "output_features": {
510
+ "action": {
511
+ "type": "ACTION",
512
+ "shape": [
513
+ 6
514
+ ]
515
+ }
516
+ },
517
+ "device": "cuda",
518
+ "use_amp": false,
519
+ "use_peft": false,
520
+ "push_to_hub": false,
521
+ "repo_id": null,
522
+ "private": null,
523
+ "tags": null,
524
+ "license": null,
525
+ "pretrained_path": null,
526
+ "chunk_size": 90,
527
+ "n_action_steps": 30,
528
+ "normalization_mapping": {
529
+ "VISUAL": "MEAN_STD",
530
+ "STATE": "MEAN_STD",
531
+ "ACTION": "MEAN_STD"
532
+ },
533
+ "vision_backbone": "resnet18",
534
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
535
+ "replace_final_stride_with_dilation": false,
536
+ "pre_norm": false,
537
+ "dim_model": 512,
538
+ "n_heads": 8,
539
+ "dim_feedforward": 3200,
540
+ "feedforward_activation": "relu",
541
+ "n_encoder_layers": 4,
542
+ "n_decoder_layers": 1,
543
+ "use_vae": true,
544
+ "latent_dim": 32,
545
+ "n_vae_encoder_layers": 4,
546
+ "temporal_ensemble_coeff": null,
547
+ "dropout": 0.1,
548
+ "kl_weight": 10.0,
549
+ "optimizer_lr": 1e-05,
550
+ "optimizer_weight_decay": 0.0001,
551
+ "optimizer_lr_backbone": 1e-05
552
+ },
553
+ "output_dir": "C:\\Users\\Usuario\\Documents\\Workspace\\LeRobot_\\demos\\06_imitation_learning\\outputs\\sweep_super",
554
+ "job_name": "act_so101_sweep_super",
555
+ "resume": false,
556
+ "seed": 1000,
557
+ "num_workers": 4,
558
+ "batch_size": 8,
559
+ "steps": 30000,
560
+ "eval_freq": 20000,
561
+ "log_freq": 100,
562
+ "tolerance_s": 0.0001,
563
+ "save_checkpoint": true,
564
+ "save_freq": 6000,
565
+ "use_policy_training_preset": true,
566
+ "optimizer": {
567
+ "type": "adamw",
568
+ "lr": 1e-05,
569
+ "weight_decay": 0.0001,
570
+ "grad_clip_norm": 10.0,
571
+ "betas": [
572
+ 0.9,
573
+ 0.999
574
+ ],
575
+ "eps": 1e-08
576
+ },
577
+ "scheduler": null,
578
+ "eval": {
579
+ "n_episodes": 50,
580
+ "batch_size": 50,
581
+ "use_async_envs": false
582
+ },
583
+ "wandb": {
584
+ "enable": false,
585
+ "disable_artifact": false,
586
+ "project": "lerobot",
587
+ "entity": null,
588
+ "notes": null,
589
+ "run_id": null,
590
+ "mode": null
591
+ },
592
+ "peft": null,
593
+ "use_rabc": false,
594
+ "rabc_progress_path": null,
595
+ "rabc_kappa": 0.01,
596
+ "rabc_epsilon": 1e-06,
597
+ "rabc_head_mode": "sparse",
598
+ "rename_map": {},
599
+ "checkpoint_path": null
600
+ }
pick_v3_super/checkpoints/018000/dataset_features.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action": {
3
+ "dtype": "float32",
4
+ "names": [
5
+ "shoulder_pan.pos",
6
+ "shoulder_lift.pos",
7
+ "elbow_flex.pos",
8
+ "wrist_flex.pos",
9
+ "wrist_roll.pos",
10
+ "gripper.pos"
11
+ ],
12
+ "shape": [
13
+ 6
14
+ ]
15
+ },
16
+ "observation.images.front": {
17
+ "dtype": "video",
18
+ "shape": [
19
+ 480,
20
+ 640,
21
+ 3
22
+ ],
23
+ "names": [
24
+ "height",
25
+ "width",
26
+ "channels"
27
+ ],
28
+ "info": {
29
+ "is_depth_map": false,
30
+ "video.height": 480,
31
+ "video.width": 640,
32
+ "video.codec": "h264",
33
+ "video.pix_fmt": "yuv420p",
34
+ "video.fps": 30,
35
+ "video.channels": 3,
36
+ "has_audio": false,
37
+ "video.g": 2,
38
+ "video.crf": 30,
39
+ "video.preset": "p1",
40
+ "video.fast_decode": 0,
41
+ "video.video_backend": "pyav",
42
+ "video.extra_options": {}
43
+ }
44
+ },
45
+ "observation.images.wrist": {
46
+ "dtype": "video",
47
+ "shape": [
48
+ 480,
49
+ 640,
50
+ 3
51
+ ],
52
+ "names": [
53
+ "height",
54
+ "width",
55
+ "channels"
56
+ ],
57
+ "info": {
58
+ "is_depth_map": false,
59
+ "video.height": 480,
60
+ "video.width": 640,
61
+ "video.codec": "h264",
62
+ "video.pix_fmt": "yuv420p",
63
+ "video.fps": 30,
64
+ "video.channels": 3,
65
+ "has_audio": false,
66
+ "video.g": 2,
67
+ "video.crf": 30,
68
+ "video.preset": "p1",
69
+ "video.fast_decode": 0,
70
+ "video.video_backend": "pyav",
71
+ "video.extra_options": {}
72
+ }
73
+ },
74
+ "observation.state": {
75
+ "dtype": "float32",
76
+ "names": [
77
+ "shoulder_pan.pos",
78
+ "shoulder_lift.pos",
79
+ "elbow_flex.pos",
80
+ "wrist_flex.pos",
81
+ "wrist_roll.pos",
82
+ "gripper.pos"
83
+ ],
84
+ "shape": [
85
+ 6
86
+ ]
87
+ }
88
+ }
pick_v3_super/checkpoints/018000/model_card_original.md ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: lerobot
4
+ pipeline_tag: robotics
5
+ tags:
6
+ - lerobot
7
+ - act
8
+ - so101
9
+ - pick
10
+ ---
11
+
12
+ # act-pick-v3-super-18000
13
+
14
+ Pick policy trained on the "super" mix of the third round (v3): every recording that passes the quality filter. Checkpoint at 18,000 steps. This is not the best checkpoint of its run (the best one is at 6,000 steps); it is published for completeness.
15
+
16
+ ACT policy for the SO-101 "pick a piece from the crate and hold it up to the camera" task. Inputs: front camera, wrist camera and the six joint positions; output: a chunk of 90 future joint targets at 30 fps.
17
+
18
+ ## Training data
19
+
20
+ 398 training episodes: all 66 pick-cube sessions, 30 fps and 10 fps (up-sampled to 30 fps), any number of pieces in the crate; 5 to 15 s long, start gap at most 20 degrees, at most 3 s idle at either end, gripper opens at least once, known bad takes removed.
21
+
22
+ Recipe: LeRobot `train()`, batch 8, chunk 90, 30 action steps per replan, default image augmentation, seed 1000.
23
+
24
+ ## Offline check on the held-out episodes
25
+
26
+ Held out: 16 episodes, 12 from `so101-pick-cube_20261024_181939` (one piece) and 4 from `so101-pick-cube_20261024_144537` (two pieces). Mean absolute joint error over the predicted chunk (lower is better) against a "stay at the current joint positions" baseline of 20.71:
27
+
28
+ | checkpoint of this run | error | below baseline |
29
+ |---|---|---|
30
+ | 6,000 steps | 15.16 | 27% |
31
+ | 12,000 steps | 16.08 | 22% |
32
+ | **18,000 steps (this one)** | **16.04** | 23% |
33
+ | 24,000 steps | 16.36 | 21% |
34
+ | 30,000 steps | 16.27 | 21% |
35
+
36
+ All runs of this round, best checkpoint of each:
37
+
38
+ | run | training episodes | best checkpoint | error | below stay-still | repo |
39
+ |---|---|---|---|---|---|
40
+ | super (all quality-passing data) | 398 | 6k | 15.16 | 27% | act-pick-v3-6000 |
41
+ | strict (6-12 s, tight quality) | 205 | 2.4k | 15.14 | 27% | act-pick-v3-strict-2400 |
42
+ | v3 recipe on 1-2 piece episodes | 331 | 18k | 15.35 | 26% | act-pick-v3-1-2-pieces-18000 |
43
+ | v2 (earlier) | 278 | 6k | 15.36 | 26% | act-pick-v2-6000 |
44
+ | chunk 60 (own baseline 16.23) | 331 | 4.8k | 11.94 | 26% | act-pick-v3-chunk60-4800 |
45
+ | one piece only | 292 | 9.6k | 15.47 | 25% | act-pick-v3-1piece-9600 |
46
+ | recent sessions only | 100 | 12k | 16.87 | 19% | act-pick-v3-recent-12000 |
47
+ | first run | 165 | 6k | 16.87 | 19% | act-pick-* |
48
+
49
+ The best checkpoints of the main runs are within noise of each other, and every run got worse after about one pass over the data. The offline error cannot say which model grasps best; none of them has been tested on the robot.
pick_v3_super/checkpoints/018000/pretrained_model/config.json ADDED
@@ -0,0 +1,71 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "act",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 6
9
+ ]
10
+ },
11
+ "observation.images.front": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 480,
16
+ 640
17
+ ]
18
+ },
19
+ "observation.images.wrist": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 480,
24
+ 640
25
+ ]
26
+ }
27
+ },
28
+ "output_features": {
29
+ "action": {
30
+ "type": "ACTION",
31
+ "shape": [
32
+ 6
33
+ ]
34
+ }
35
+ },
36
+ "device": "cuda",
37
+ "use_amp": false,
38
+ "use_peft": false,
39
+ "push_to_hub": false,
40
+ "repo_id": null,
41
+ "private": null,
42
+ "tags": null,
43
+ "license": null,
44
+ "pretrained_path": null,
45
+ "chunk_size": 90,
46
+ "n_action_steps": 30,
47
+ "normalization_mapping": {
48
+ "VISUAL": "MEAN_STD",
49
+ "STATE": "MEAN_STD",
50
+ "ACTION": "MEAN_STD"
51
+ },
52
+ "vision_backbone": "resnet18",
53
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
54
+ "replace_final_stride_with_dilation": false,
55
+ "pre_norm": false,
56
+ "dim_model": 512,
57
+ "n_heads": 8,
58
+ "dim_feedforward": 3200,
59
+ "feedforward_activation": "relu",
60
+ "n_encoder_layers": 4,
61
+ "n_decoder_layers": 1,
62
+ "use_vae": true,
63
+ "latent_dim": 32,
64
+ "n_vae_encoder_layers": 4,
65
+ "temporal_ensemble_coeff": null,
66
+ "dropout": 0.1,
67
+ "kl_weight": 10.0,
68
+ "optimizer_lr": 1e-05,
69
+ "optimizer_weight_decay": 0.0001,
70
+ "optimizer_lr_backbone": 1e-05
71
+ }
pick_v3_super/checkpoints/018000/pretrained_model/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:762d769e8e75541bde09148f6e30be4bb92ab40fb529f6a254832f07c7f3e7f4
3
+ size 206658768
pick_v3_super/checkpoints/018000/pretrained_model/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
+ 6
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
+ }
pick_v3_super/checkpoints/018000/pretrained_model/policy_postprocessor_step_0_unnormalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931ed929c95ef00f6e334882511164d4c9d89f16bfb31eeb212ddecf99897ed7
3
+ size 7552
pick_v3_super/checkpoints/018000/pretrained_model/policy_preprocessor.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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.state": {
27
+ "type": "STATE",
28
+ "shape": [
29
+ 6
30
+ ]
31
+ },
32
+ "observation.images.front": {
33
+ "type": "VISUAL",
34
+ "shape": [
35
+ 3,
36
+ 480,
37
+ 640
38
+ ]
39
+ },
40
+ "observation.images.wrist": {
41
+ "type": "VISUAL",
42
+ "shape": [
43
+ 3,
44
+ 480,
45
+ 640
46
+ ]
47
+ },
48
+ "action": {
49
+ "type": "ACTION",
50
+ "shape": [
51
+ 6
52
+ ]
53
+ }
54
+ },
55
+ "norm_map": {
56
+ "VISUAL": "MEAN_STD",
57
+ "STATE": "MEAN_STD",
58
+ "ACTION": "MEAN_STD"
59
+ }
60
+ },
61
+ "state_file": "policy_preprocessor_step_3_normalizer_processor.safetensors"
62
+ }
63
+ ]
64
+ }
pick_v3_super/checkpoints/018000/pretrained_model/policy_preprocessor_step_3_normalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931ed929c95ef00f6e334882511164d4c9d89f16bfb31eeb212ddecf99897ed7
3
+ size 7552
pick_v3_super/checkpoints/018000/pretrained_model/train_config.json ADDED
@@ -0,0 +1,600 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "local/so101_act",
4
+ "root": "C:\\Users\\Usuario\\Documents\\Workspace\\LeRobot_\\demos\\06_imitation_learning\\data\\real_so101\\act_pick_super",
5
+ "episodes": [
6
+ 4,
7
+ 5,
8
+ 6,
9
+ 7,
10
+ 8,
11
+ 9,
12
+ 10,
13
+ 12,
14
+ 15,
15
+ 17,
16
+ 18,
17
+ 19,
18
+ 20,
19
+ 21,
20
+ 23,
21
+ 24,
22
+ 25,
23
+ 27,
24
+ 28,
25
+ 29,
26
+ 30,
27
+ 31,
28
+ 32,
29
+ 33,
30
+ 34,
31
+ 35,
32
+ 36,
33
+ 37,
34
+ 38,
35
+ 39,
36
+ 41,
37
+ 42,
38
+ 43,
39
+ 44,
40
+ 45,
41
+ 46,
42
+ 47,
43
+ 49,
44
+ 50,
45
+ 51,
46
+ 52,
47
+ 53,
48
+ 54,
49
+ 55,
50
+ 56,
51
+ 57,
52
+ 58,
53
+ 59,
54
+ 60,
55
+ 63,
56
+ 64,
57
+ 65,
58
+ 66,
59
+ 67,
60
+ 68,
61
+ 69,
62
+ 70,
63
+ 71,
64
+ 72,
65
+ 73,
66
+ 74,
67
+ 76,
68
+ 77,
69
+ 78,
70
+ 79,
71
+ 84,
72
+ 86,
73
+ 87,
74
+ 88,
75
+ 89,
76
+ 90,
77
+ 91,
78
+ 92,
79
+ 93,
80
+ 94,
81
+ 96,
82
+ 97,
83
+ 98,
84
+ 100,
85
+ 101,
86
+ 102,
87
+ 103,
88
+ 104,
89
+ 105,
90
+ 106,
91
+ 107,
92
+ 109,
93
+ 110,
94
+ 111,
95
+ 112,
96
+ 113,
97
+ 114,
98
+ 115,
99
+ 116,
100
+ 117,
101
+ 118,
102
+ 119,
103
+ 120,
104
+ 121,
105
+ 122,
106
+ 124,
107
+ 125,
108
+ 126,
109
+ 127,
110
+ 128,
111
+ 129,
112
+ 130,
113
+ 131,
114
+ 132,
115
+ 133,
116
+ 134,
117
+ 135,
118
+ 136,
119
+ 137,
120
+ 138,
121
+ 139,
122
+ 140,
123
+ 141,
124
+ 142,
125
+ 143,
126
+ 144,
127
+ 145,
128
+ 146,
129
+ 147,
130
+ 148,
131
+ 149,
132
+ 150,
133
+ 151,
134
+ 152,
135
+ 153,
136
+ 154,
137
+ 155,
138
+ 156,
139
+ 157,
140
+ 158,
141
+ 159,
142
+ 160,
143
+ 161,
144
+ 162,
145
+ 163,
146
+ 164,
147
+ 165,
148
+ 166,
149
+ 167,
150
+ 168,
151
+ 169,
152
+ 170,
153
+ 171,
154
+ 172,
155
+ 173,
156
+ 174,
157
+ 175,
158
+ 176,
159
+ 177,
160
+ 178,
161
+ 179,
162
+ 180,
163
+ 181,
164
+ 182,
165
+ 183,
166
+ 184,
167
+ 185,
168
+ 186,
169
+ 187,
170
+ 188,
171
+ 189,
172
+ 190,
173
+ 191,
174
+ 192,
175
+ 193,
176
+ 194,
177
+ 195,
178
+ 196,
179
+ 197,
180
+ 198,
181
+ 199,
182
+ 200,
183
+ 201,
184
+ 202,
185
+ 203,
186
+ 204,
187
+ 205,
188
+ 206,
189
+ 207,
190
+ 208,
191
+ 209,
192
+ 210,
193
+ 211,
194
+ 212,
195
+ 213,
196
+ 214,
197
+ 215,
198
+ 216,
199
+ 217,
200
+ 218,
201
+ 219,
202
+ 220,
203
+ 221,
204
+ 222,
205
+ 223,
206
+ 224,
207
+ 225,
208
+ 226,
209
+ 227,
210
+ 228,
211
+ 229,
212
+ 230,
213
+ 231,
214
+ 232,
215
+ 233,
216
+ 235,
217
+ 236,
218
+ 237,
219
+ 238,
220
+ 239,
221
+ 240,
222
+ 241,
223
+ 242,
224
+ 243,
225
+ 244,
226
+ 245,
227
+ 246,
228
+ 247,
229
+ 248,
230
+ 249,
231
+ 250,
232
+ 251,
233
+ 252,
234
+ 253,
235
+ 255,
236
+ 256,
237
+ 257,
238
+ 258,
239
+ 259,
240
+ 260,
241
+ 261,
242
+ 262,
243
+ 263,
244
+ 264,
245
+ 265,
246
+ 266,
247
+ 267,
248
+ 268,
249
+ 269,
250
+ 270,
251
+ 271,
252
+ 272,
253
+ 273,
254
+ 274,
255
+ 275,
256
+ 276,
257
+ 277,
258
+ 278,
259
+ 279,
260
+ 280,
261
+ 281,
262
+ 282,
263
+ 283,
264
+ 284,
265
+ 285,
266
+ 286,
267
+ 287,
268
+ 288,
269
+ 289,
270
+ 290,
271
+ 291,
272
+ 292,
273
+ 293,
274
+ 294,
275
+ 295,
276
+ 296,
277
+ 297,
278
+ 298,
279
+ 299,
280
+ 300,
281
+ 301,
282
+ 302,
283
+ 303,
284
+ 304,
285
+ 305,
286
+ 306,
287
+ 307,
288
+ 308,
289
+ 309,
290
+ 310,
291
+ 311,
292
+ 312,
293
+ 313,
294
+ 314,
295
+ 315,
296
+ 316,
297
+ 317,
298
+ 318,
299
+ 319,
300
+ 320,
301
+ 325,
302
+ 326,
303
+ 327,
304
+ 329,
305
+ 330,
306
+ 331,
307
+ 332,
308
+ 333,
309
+ 334,
310
+ 335,
311
+ 336,
312
+ 337,
313
+ 338,
314
+ 339,
315
+ 340,
316
+ 341,
317
+ 342,
318
+ 343,
319
+ 344,
320
+ 345,
321
+ 346,
322
+ 347,
323
+ 348,
324
+ 349,
325
+ 351,
326
+ 352,
327
+ 353,
328
+ 354,
329
+ 355,
330
+ 356,
331
+ 357,
332
+ 358,
333
+ 359,
334
+ 360,
335
+ 361,
336
+ 364,
337
+ 365,
338
+ 366,
339
+ 367,
340
+ 368,
341
+ 369,
342
+ 370,
343
+ 371,
344
+ 372,
345
+ 373,
346
+ 374,
347
+ 375,
348
+ 376,
349
+ 377,
350
+ 378,
351
+ 391,
352
+ 392,
353
+ 393,
354
+ 394,
355
+ 395,
356
+ 396,
357
+ 397,
358
+ 398,
359
+ 399,
360
+ 400,
361
+ 401,
362
+ 402,
363
+ 403,
364
+ 404,
365
+ 405,
366
+ 406,
367
+ 407,
368
+ 408,
369
+ 409,
370
+ 410,
371
+ 411,
372
+ 412,
373
+ 413,
374
+ 414,
375
+ 415,
376
+ 416,
377
+ 417,
378
+ 419,
379
+ 420,
380
+ 421,
381
+ 422,
382
+ 423,
383
+ 424,
384
+ 425,
385
+ 426,
386
+ 427,
387
+ 428,
388
+ 429,
389
+ 430,
390
+ 431,
391
+ 432,
392
+ 433,
393
+ 434,
394
+ 435,
395
+ 436,
396
+ 438,
397
+ 439,
398
+ 440,
399
+ 441,
400
+ 442,
401
+ 443,
402
+ 444,
403
+ 445
404
+ ],
405
+ "image_transforms": {
406
+ "enable": true,
407
+ "max_num_transforms": 3,
408
+ "random_order": false,
409
+ "tfs": {
410
+ "brightness": {
411
+ "weight": 1.0,
412
+ "type": "ColorJitter",
413
+ "kwargs": {
414
+ "brightness": [
415
+ 0.8,
416
+ 1.2
417
+ ]
418
+ }
419
+ },
420
+ "contrast": {
421
+ "weight": 1.0,
422
+ "type": "ColorJitter",
423
+ "kwargs": {
424
+ "contrast": [
425
+ 0.8,
426
+ 1.2
427
+ ]
428
+ }
429
+ },
430
+ "saturation": {
431
+ "weight": 1.0,
432
+ "type": "ColorJitter",
433
+ "kwargs": {
434
+ "saturation": [
435
+ 0.5,
436
+ 1.5
437
+ ]
438
+ }
439
+ },
440
+ "hue": {
441
+ "weight": 1.0,
442
+ "type": "ColorJitter",
443
+ "kwargs": {
444
+ "hue": [
445
+ -0.05,
446
+ 0.05
447
+ ]
448
+ }
449
+ },
450
+ "sharpness": {
451
+ "weight": 1.0,
452
+ "type": "SharpnessJitter",
453
+ "kwargs": {
454
+ "sharpness": [
455
+ 0.5,
456
+ 1.5
457
+ ]
458
+ }
459
+ },
460
+ "affine": {
461
+ "weight": 1.0,
462
+ "type": "RandomAffine",
463
+ "kwargs": {
464
+ "degrees": [
465
+ -5.0,
466
+ 5.0
467
+ ],
468
+ "translate": [
469
+ 0.05,
470
+ 0.05
471
+ ]
472
+ }
473
+ }
474
+ }
475
+ },
476
+ "revision": null,
477
+ "use_imagenet_stats": true,
478
+ "video_backend": "pyav",
479
+ "streaming": false
480
+ },
481
+ "env": null,
482
+ "policy": {
483
+ "type": "act",
484
+ "n_obs_steps": 1,
485
+ "input_features": {
486
+ "observation.state": {
487
+ "type": "STATE",
488
+ "shape": [
489
+ 6
490
+ ]
491
+ },
492
+ "observation.images.front": {
493
+ "type": "VISUAL",
494
+ "shape": [
495
+ 3,
496
+ 480,
497
+ 640
498
+ ]
499
+ },
500
+ "observation.images.wrist": {
501
+ "type": "VISUAL",
502
+ "shape": [
503
+ 3,
504
+ 480,
505
+ 640
506
+ ]
507
+ }
508
+ },
509
+ "output_features": {
510
+ "action": {
511
+ "type": "ACTION",
512
+ "shape": [
513
+ 6
514
+ ]
515
+ }
516
+ },
517
+ "device": "cuda",
518
+ "use_amp": false,
519
+ "use_peft": false,
520
+ "push_to_hub": false,
521
+ "repo_id": null,
522
+ "private": null,
523
+ "tags": null,
524
+ "license": null,
525
+ "pretrained_path": null,
526
+ "chunk_size": 90,
527
+ "n_action_steps": 30,
528
+ "normalization_mapping": {
529
+ "VISUAL": "MEAN_STD",
530
+ "STATE": "MEAN_STD",
531
+ "ACTION": "MEAN_STD"
532
+ },
533
+ "vision_backbone": "resnet18",
534
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
535
+ "replace_final_stride_with_dilation": false,
536
+ "pre_norm": false,
537
+ "dim_model": 512,
538
+ "n_heads": 8,
539
+ "dim_feedforward": 3200,
540
+ "feedforward_activation": "relu",
541
+ "n_encoder_layers": 4,
542
+ "n_decoder_layers": 1,
543
+ "use_vae": true,
544
+ "latent_dim": 32,
545
+ "n_vae_encoder_layers": 4,
546
+ "temporal_ensemble_coeff": null,
547
+ "dropout": 0.1,
548
+ "kl_weight": 10.0,
549
+ "optimizer_lr": 1e-05,
550
+ "optimizer_weight_decay": 0.0001,
551
+ "optimizer_lr_backbone": 1e-05
552
+ },
553
+ "output_dir": "C:\\Users\\Usuario\\Documents\\Workspace\\LeRobot_\\demos\\06_imitation_learning\\outputs\\sweep_super",
554
+ "job_name": "act_so101_sweep_super",
555
+ "resume": false,
556
+ "seed": 1000,
557
+ "num_workers": 4,
558
+ "batch_size": 8,
559
+ "steps": 30000,
560
+ "eval_freq": 20000,
561
+ "log_freq": 100,
562
+ "tolerance_s": 0.0001,
563
+ "save_checkpoint": true,
564
+ "save_freq": 6000,
565
+ "use_policy_training_preset": true,
566
+ "optimizer": {
567
+ "type": "adamw",
568
+ "lr": 1e-05,
569
+ "weight_decay": 0.0001,
570
+ "grad_clip_norm": 10.0,
571
+ "betas": [
572
+ 0.9,
573
+ 0.999
574
+ ],
575
+ "eps": 1e-08
576
+ },
577
+ "scheduler": null,
578
+ "eval": {
579
+ "n_episodes": 50,
580
+ "batch_size": 50,
581
+ "use_async_envs": false
582
+ },
583
+ "wandb": {
584
+ "enable": false,
585
+ "disable_artifact": false,
586
+ "project": "lerobot",
587
+ "entity": null,
588
+ "notes": null,
589
+ "run_id": null,
590
+ "mode": null
591
+ },
592
+ "peft": null,
593
+ "use_rabc": false,
594
+ "rabc_progress_path": null,
595
+ "rabc_kappa": 0.01,
596
+ "rabc_epsilon": 1e-06,
597
+ "rabc_head_mode": "sparse",
598
+ "rename_map": {},
599
+ "checkpoint_path": null
600
+ }
pick_v3_super/checkpoints/024000/dataset_features.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action": {
3
+ "dtype": "float32",
4
+ "names": [
5
+ "shoulder_pan.pos",
6
+ "shoulder_lift.pos",
7
+ "elbow_flex.pos",
8
+ "wrist_flex.pos",
9
+ "wrist_roll.pos",
10
+ "gripper.pos"
11
+ ],
12
+ "shape": [
13
+ 6
14
+ ]
15
+ },
16
+ "observation.images.front": {
17
+ "dtype": "video",
18
+ "shape": [
19
+ 480,
20
+ 640,
21
+ 3
22
+ ],
23
+ "names": [
24
+ "height",
25
+ "width",
26
+ "channels"
27
+ ],
28
+ "info": {
29
+ "is_depth_map": false,
30
+ "video.height": 480,
31
+ "video.width": 640,
32
+ "video.codec": "h264",
33
+ "video.pix_fmt": "yuv420p",
34
+ "video.fps": 30,
35
+ "video.channels": 3,
36
+ "has_audio": false,
37
+ "video.g": 2,
38
+ "video.crf": 30,
39
+ "video.preset": "p1",
40
+ "video.fast_decode": 0,
41
+ "video.video_backend": "pyav",
42
+ "video.extra_options": {}
43
+ }
44
+ },
45
+ "observation.images.wrist": {
46
+ "dtype": "video",
47
+ "shape": [
48
+ 480,
49
+ 640,
50
+ 3
51
+ ],
52
+ "names": [
53
+ "height",
54
+ "width",
55
+ "channels"
56
+ ],
57
+ "info": {
58
+ "is_depth_map": false,
59
+ "video.height": 480,
60
+ "video.width": 640,
61
+ "video.codec": "h264",
62
+ "video.pix_fmt": "yuv420p",
63
+ "video.fps": 30,
64
+ "video.channels": 3,
65
+ "has_audio": false,
66
+ "video.g": 2,
67
+ "video.crf": 30,
68
+ "video.preset": "p1",
69
+ "video.fast_decode": 0,
70
+ "video.video_backend": "pyav",
71
+ "video.extra_options": {}
72
+ }
73
+ },
74
+ "observation.state": {
75
+ "dtype": "float32",
76
+ "names": [
77
+ "shoulder_pan.pos",
78
+ "shoulder_lift.pos",
79
+ "elbow_flex.pos",
80
+ "wrist_flex.pos",
81
+ "wrist_roll.pos",
82
+ "gripper.pos"
83
+ ],
84
+ "shape": [
85
+ 6
86
+ ]
87
+ }
88
+ }
pick_v3_super/checkpoints/024000/model_card_original.md ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: lerobot
4
+ pipeline_tag: robotics
5
+ tags:
6
+ - lerobot
7
+ - act
8
+ - so101
9
+ - pick
10
+ ---
11
+
12
+ # act-pick-v3-super-24000
13
+
14
+ Pick policy trained on the "super" mix of the third round (v3): every recording that passes the quality filter. Checkpoint at 24,000 steps. This is not the best checkpoint of its run (the best one is at 6,000 steps); it is published for completeness.
15
+
16
+ ACT policy for the SO-101 "pick a piece from the crate and hold it up to the camera" task. Inputs: front camera, wrist camera and the six joint positions; output: a chunk of 90 future joint targets at 30 fps.
17
+
18
+ ## Training data
19
+
20
+ 398 training episodes: all 66 pick-cube sessions, 30 fps and 10 fps (up-sampled to 30 fps), any number of pieces in the crate; 5 to 15 s long, start gap at most 20 degrees, at most 3 s idle at either end, gripper opens at least once, known bad takes removed.
21
+
22
+ Recipe: LeRobot `train()`, batch 8, chunk 90, 30 action steps per replan, default image augmentation, seed 1000.
23
+
24
+ ## Offline check on the held-out episodes
25
+
26
+ Held out: 16 episodes, 12 from `so101-pick-cube_20261024_181939` (one piece) and 4 from `so101-pick-cube_20261024_144537` (two pieces). Mean absolute joint error over the predicted chunk (lower is better) against a "stay at the current joint positions" baseline of 20.71:
27
+
28
+ | checkpoint of this run | error | below baseline |
29
+ |---|---|---|
30
+ | 6,000 steps | 15.16 | 27% |
31
+ | 12,000 steps | 16.08 | 22% |
32
+ | 18,000 steps | 16.04 | 23% |
33
+ | **24,000 steps (this one)** | **16.36** | 21% |
34
+ | 30,000 steps | 16.27 | 21% |
35
+
36
+ All runs of this round, best checkpoint of each:
37
+
38
+ | run | training episodes | best checkpoint | error | below stay-still | repo |
39
+ |---|---|---|---|---|---|
40
+ | super (all quality-passing data) | 398 | 6k | 15.16 | 27% | act-pick-v3-6000 |
41
+ | strict (6-12 s, tight quality) | 205 | 2.4k | 15.14 | 27% | act-pick-v3-strict-2400 |
42
+ | v3 recipe on 1-2 piece episodes | 331 | 18k | 15.35 | 26% | act-pick-v3-1-2-pieces-18000 |
43
+ | v2 (earlier) | 278 | 6k | 15.36 | 26% | act-pick-v2-6000 |
44
+ | chunk 60 (own baseline 16.23) | 331 | 4.8k | 11.94 | 26% | act-pick-v3-chunk60-4800 |
45
+ | one piece only | 292 | 9.6k | 15.47 | 25% | act-pick-v3-1piece-9600 |
46
+ | recent sessions only | 100 | 12k | 16.87 | 19% | act-pick-v3-recent-12000 |
47
+ | first run | 165 | 6k | 16.87 | 19% | act-pick-* |
48
+
49
+ The best checkpoints of the main runs are within noise of each other, and every run got worse after about one pass over the data. The offline error cannot say which model grasps best; none of them has been tested on the robot.
pick_v3_super/checkpoints/024000/pretrained_model/config.json ADDED
@@ -0,0 +1,71 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "act",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 6
9
+ ]
10
+ },
11
+ "observation.images.front": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 480,
16
+ 640
17
+ ]
18
+ },
19
+ "observation.images.wrist": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 480,
24
+ 640
25
+ ]
26
+ }
27
+ },
28
+ "output_features": {
29
+ "action": {
30
+ "type": "ACTION",
31
+ "shape": [
32
+ 6
33
+ ]
34
+ }
35
+ },
36
+ "device": "cuda",
37
+ "use_amp": false,
38
+ "use_peft": false,
39
+ "push_to_hub": false,
40
+ "repo_id": null,
41
+ "private": null,
42
+ "tags": null,
43
+ "license": null,
44
+ "pretrained_path": null,
45
+ "chunk_size": 90,
46
+ "n_action_steps": 30,
47
+ "normalization_mapping": {
48
+ "VISUAL": "MEAN_STD",
49
+ "STATE": "MEAN_STD",
50
+ "ACTION": "MEAN_STD"
51
+ },
52
+ "vision_backbone": "resnet18",
53
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
54
+ "replace_final_stride_with_dilation": false,
55
+ "pre_norm": false,
56
+ "dim_model": 512,
57
+ "n_heads": 8,
58
+ "dim_feedforward": 3200,
59
+ "feedforward_activation": "relu",
60
+ "n_encoder_layers": 4,
61
+ "n_decoder_layers": 1,
62
+ "use_vae": true,
63
+ "latent_dim": 32,
64
+ "n_vae_encoder_layers": 4,
65
+ "temporal_ensemble_coeff": null,
66
+ "dropout": 0.1,
67
+ "kl_weight": 10.0,
68
+ "optimizer_lr": 1e-05,
69
+ "optimizer_weight_decay": 0.0001,
70
+ "optimizer_lr_backbone": 1e-05
71
+ }
pick_v3_super/checkpoints/024000/pretrained_model/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59ab69e9f8b57623b916c48904b1ebbac755ee090223a0e3dd047fdd5d09e315
3
+ size 206658768
pick_v3_super/checkpoints/024000/pretrained_model/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
+ 6
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
+ }
pick_v3_super/checkpoints/024000/pretrained_model/policy_postprocessor_step_0_unnormalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931ed929c95ef00f6e334882511164d4c9d89f16bfb31eeb212ddecf99897ed7
3
+ size 7552
pick_v3_super/checkpoints/024000/pretrained_model/policy_preprocessor.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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.state": {
27
+ "type": "STATE",
28
+ "shape": [
29
+ 6
30
+ ]
31
+ },
32
+ "observation.images.front": {
33
+ "type": "VISUAL",
34
+ "shape": [
35
+ 3,
36
+ 480,
37
+ 640
38
+ ]
39
+ },
40
+ "observation.images.wrist": {
41
+ "type": "VISUAL",
42
+ "shape": [
43
+ 3,
44
+ 480,
45
+ 640
46
+ ]
47
+ },
48
+ "action": {
49
+ "type": "ACTION",
50
+ "shape": [
51
+ 6
52
+ ]
53
+ }
54
+ },
55
+ "norm_map": {
56
+ "VISUAL": "MEAN_STD",
57
+ "STATE": "MEAN_STD",
58
+ "ACTION": "MEAN_STD"
59
+ }
60
+ },
61
+ "state_file": "policy_preprocessor_step_3_normalizer_processor.safetensors"
62
+ }
63
+ ]
64
+ }
pick_v3_super/checkpoints/024000/pretrained_model/policy_preprocessor_step_3_normalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931ed929c95ef00f6e334882511164d4c9d89f16bfb31eeb212ddecf99897ed7
3
+ size 7552
pick_v3_super/checkpoints/024000/pretrained_model/train_config.json ADDED
@@ -0,0 +1,600 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "local/so101_act",
4
+ "root": "C:\\Users\\Usuario\\Documents\\Workspace\\LeRobot_\\demos\\06_imitation_learning\\data\\real_so101\\act_pick_super",
5
+ "episodes": [
6
+ 4,
7
+ 5,
8
+ 6,
9
+ 7,
10
+ 8,
11
+ 9,
12
+ 10,
13
+ 12,
14
+ 15,
15
+ 17,
16
+ 18,
17
+ 19,
18
+ 20,
19
+ 21,
20
+ 23,
21
+ 24,
22
+ 25,
23
+ 27,
24
+ 28,
25
+ 29,
26
+ 30,
27
+ 31,
28
+ 32,
29
+ 33,
30
+ 34,
31
+ 35,
32
+ 36,
33
+ 37,
34
+ 38,
35
+ 39,
36
+ 41,
37
+ 42,
38
+ 43,
39
+ 44,
40
+ 45,
41
+ 46,
42
+ 47,
43
+ 49,
44
+ 50,
45
+ 51,
46
+ 52,
47
+ 53,
48
+ 54,
49
+ 55,
50
+ 56,
51
+ 57,
52
+ 58,
53
+ 59,
54
+ 60,
55
+ 63,
56
+ 64,
57
+ 65,
58
+ 66,
59
+ 67,
60
+ 68,
61
+ 69,
62
+ 70,
63
+ 71,
64
+ 72,
65
+ 73,
66
+ 74,
67
+ 76,
68
+ 77,
69
+ 78,
70
+ 79,
71
+ 84,
72
+ 86,
73
+ 87,
74
+ 88,
75
+ 89,
76
+ 90,
77
+ 91,
78
+ 92,
79
+ 93,
80
+ 94,
81
+ 96,
82
+ 97,
83
+ 98,
84
+ 100,
85
+ 101,
86
+ 102,
87
+ 103,
88
+ 104,
89
+ 105,
90
+ 106,
91
+ 107,
92
+ 109,
93
+ 110,
94
+ 111,
95
+ 112,
96
+ 113,
97
+ 114,
98
+ 115,
99
+ 116,
100
+ 117,
101
+ 118,
102
+ 119,
103
+ 120,
104
+ 121,
105
+ 122,
106
+ 124,
107
+ 125,
108
+ 126,
109
+ 127,
110
+ 128,
111
+ 129,
112
+ 130,
113
+ 131,
114
+ 132,
115
+ 133,
116
+ 134,
117
+ 135,
118
+ 136,
119
+ 137,
120
+ 138,
121
+ 139,
122
+ 140,
123
+ 141,
124
+ 142,
125
+ 143,
126
+ 144,
127
+ 145,
128
+ 146,
129
+ 147,
130
+ 148,
131
+ 149,
132
+ 150,
133
+ 151,
134
+ 152,
135
+ 153,
136
+ 154,
137
+ 155,
138
+ 156,
139
+ 157,
140
+ 158,
141
+ 159,
142
+ 160,
143
+ 161,
144
+ 162,
145
+ 163,
146
+ 164,
147
+ 165,
148
+ 166,
149
+ 167,
150
+ 168,
151
+ 169,
152
+ 170,
153
+ 171,
154
+ 172,
155
+ 173,
156
+ 174,
157
+ 175,
158
+ 176,
159
+ 177,
160
+ 178,
161
+ 179,
162
+ 180,
163
+ 181,
164
+ 182,
165
+ 183,
166
+ 184,
167
+ 185,
168
+ 186,
169
+ 187,
170
+ 188,
171
+ 189,
172
+ 190,
173
+ 191,
174
+ 192,
175
+ 193,
176
+ 194,
177
+ 195,
178
+ 196,
179
+ 197,
180
+ 198,
181
+ 199,
182
+ 200,
183
+ 201,
184
+ 202,
185
+ 203,
186
+ 204,
187
+ 205,
188
+ 206,
189
+ 207,
190
+ 208,
191
+ 209,
192
+ 210,
193
+ 211,
194
+ 212,
195
+ 213,
196
+ 214,
197
+ 215,
198
+ 216,
199
+ 217,
200
+ 218,
201
+ 219,
202
+ 220,
203
+ 221,
204
+ 222,
205
+ 223,
206
+ 224,
207
+ 225,
208
+ 226,
209
+ 227,
210
+ 228,
211
+ 229,
212
+ 230,
213
+ 231,
214
+ 232,
215
+ 233,
216
+ 235,
217
+ 236,
218
+ 237,
219
+ 238,
220
+ 239,
221
+ 240,
222
+ 241,
223
+ 242,
224
+ 243,
225
+ 244,
226
+ 245,
227
+ 246,
228
+ 247,
229
+ 248,
230
+ 249,
231
+ 250,
232
+ 251,
233
+ 252,
234
+ 253,
235
+ 255,
236
+ 256,
237
+ 257,
238
+ 258,
239
+ 259,
240
+ 260,
241
+ 261,
242
+ 262,
243
+ 263,
244
+ 264,
245
+ 265,
246
+ 266,
247
+ 267,
248
+ 268,
249
+ 269,
250
+ 270,
251
+ 271,
252
+ 272,
253
+ 273,
254
+ 274,
255
+ 275,
256
+ 276,
257
+ 277,
258
+ 278,
259
+ 279,
260
+ 280,
261
+ 281,
262
+ 282,
263
+ 283,
264
+ 284,
265
+ 285,
266
+ 286,
267
+ 287,
268
+ 288,
269
+ 289,
270
+ 290,
271
+ 291,
272
+ 292,
273
+ 293,
274
+ 294,
275
+ 295,
276
+ 296,
277
+ 297,
278
+ 298,
279
+ 299,
280
+ 300,
281
+ 301,
282
+ 302,
283
+ 303,
284
+ 304,
285
+ 305,
286
+ 306,
287
+ 307,
288
+ 308,
289
+ 309,
290
+ 310,
291
+ 311,
292
+ 312,
293
+ 313,
294
+ 314,
295
+ 315,
296
+ 316,
297
+ 317,
298
+ 318,
299
+ 319,
300
+ 320,
301
+ 325,
302
+ 326,
303
+ 327,
304
+ 329,
305
+ 330,
306
+ 331,
307
+ 332,
308
+ 333,
309
+ 334,
310
+ 335,
311
+ 336,
312
+ 337,
313
+ 338,
314
+ 339,
315
+ 340,
316
+ 341,
317
+ 342,
318
+ 343,
319
+ 344,
320
+ 345,
321
+ 346,
322
+ 347,
323
+ 348,
324
+ 349,
325
+ 351,
326
+ 352,
327
+ 353,
328
+ 354,
329
+ 355,
330
+ 356,
331
+ 357,
332
+ 358,
333
+ 359,
334
+ 360,
335
+ 361,
336
+ 364,
337
+ 365,
338
+ 366,
339
+ 367,
340
+ 368,
341
+ 369,
342
+ 370,
343
+ 371,
344
+ 372,
345
+ 373,
346
+ 374,
347
+ 375,
348
+ 376,
349
+ 377,
350
+ 378,
351
+ 391,
352
+ 392,
353
+ 393,
354
+ 394,
355
+ 395,
356
+ 396,
357
+ 397,
358
+ 398,
359
+ 399,
360
+ 400,
361
+ 401,
362
+ 402,
363
+ 403,
364
+ 404,
365
+ 405,
366
+ 406,
367
+ 407,
368
+ 408,
369
+ 409,
370
+ 410,
371
+ 411,
372
+ 412,
373
+ 413,
374
+ 414,
375
+ 415,
376
+ 416,
377
+ 417,
378
+ 419,
379
+ 420,
380
+ 421,
381
+ 422,
382
+ 423,
383
+ 424,
384
+ 425,
385
+ 426,
386
+ 427,
387
+ 428,
388
+ 429,
389
+ 430,
390
+ 431,
391
+ 432,
392
+ 433,
393
+ 434,
394
+ 435,
395
+ 436,
396
+ 438,
397
+ 439,
398
+ 440,
399
+ 441,
400
+ 442,
401
+ 443,
402
+ 444,
403
+ 445
404
+ ],
405
+ "image_transforms": {
406
+ "enable": true,
407
+ "max_num_transforms": 3,
408
+ "random_order": false,
409
+ "tfs": {
410
+ "brightness": {
411
+ "weight": 1.0,
412
+ "type": "ColorJitter",
413
+ "kwargs": {
414
+ "brightness": [
415
+ 0.8,
416
+ 1.2
417
+ ]
418
+ }
419
+ },
420
+ "contrast": {
421
+ "weight": 1.0,
422
+ "type": "ColorJitter",
423
+ "kwargs": {
424
+ "contrast": [
425
+ 0.8,
426
+ 1.2
427
+ ]
428
+ }
429
+ },
430
+ "saturation": {
431
+ "weight": 1.0,
432
+ "type": "ColorJitter",
433
+ "kwargs": {
434
+ "saturation": [
435
+ 0.5,
436
+ 1.5
437
+ ]
438
+ }
439
+ },
440
+ "hue": {
441
+ "weight": 1.0,
442
+ "type": "ColorJitter",
443
+ "kwargs": {
444
+ "hue": [
445
+ -0.05,
446
+ 0.05
447
+ ]
448
+ }
449
+ },
450
+ "sharpness": {
451
+ "weight": 1.0,
452
+ "type": "SharpnessJitter",
453
+ "kwargs": {
454
+ "sharpness": [
455
+ 0.5,
456
+ 1.5
457
+ ]
458
+ }
459
+ },
460
+ "affine": {
461
+ "weight": 1.0,
462
+ "type": "RandomAffine",
463
+ "kwargs": {
464
+ "degrees": [
465
+ -5.0,
466
+ 5.0
467
+ ],
468
+ "translate": [
469
+ 0.05,
470
+ 0.05
471
+ ]
472
+ }
473
+ }
474
+ }
475
+ },
476
+ "revision": null,
477
+ "use_imagenet_stats": true,
478
+ "video_backend": "pyav",
479
+ "streaming": false
480
+ },
481
+ "env": null,
482
+ "policy": {
483
+ "type": "act",
484
+ "n_obs_steps": 1,
485
+ "input_features": {
486
+ "observation.state": {
487
+ "type": "STATE",
488
+ "shape": [
489
+ 6
490
+ ]
491
+ },
492
+ "observation.images.front": {
493
+ "type": "VISUAL",
494
+ "shape": [
495
+ 3,
496
+ 480,
497
+ 640
498
+ ]
499
+ },
500
+ "observation.images.wrist": {
501
+ "type": "VISUAL",
502
+ "shape": [
503
+ 3,
504
+ 480,
505
+ 640
506
+ ]
507
+ }
508
+ },
509
+ "output_features": {
510
+ "action": {
511
+ "type": "ACTION",
512
+ "shape": [
513
+ 6
514
+ ]
515
+ }
516
+ },
517
+ "device": "cuda",
518
+ "use_amp": false,
519
+ "use_peft": false,
520
+ "push_to_hub": false,
521
+ "repo_id": null,
522
+ "private": null,
523
+ "tags": null,
524
+ "license": null,
525
+ "pretrained_path": null,
526
+ "chunk_size": 90,
527
+ "n_action_steps": 30,
528
+ "normalization_mapping": {
529
+ "VISUAL": "MEAN_STD",
530
+ "STATE": "MEAN_STD",
531
+ "ACTION": "MEAN_STD"
532
+ },
533
+ "vision_backbone": "resnet18",
534
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
535
+ "replace_final_stride_with_dilation": false,
536
+ "pre_norm": false,
537
+ "dim_model": 512,
538
+ "n_heads": 8,
539
+ "dim_feedforward": 3200,
540
+ "feedforward_activation": "relu",
541
+ "n_encoder_layers": 4,
542
+ "n_decoder_layers": 1,
543
+ "use_vae": true,
544
+ "latent_dim": 32,
545
+ "n_vae_encoder_layers": 4,
546
+ "temporal_ensemble_coeff": null,
547
+ "dropout": 0.1,
548
+ "kl_weight": 10.0,
549
+ "optimizer_lr": 1e-05,
550
+ "optimizer_weight_decay": 0.0001,
551
+ "optimizer_lr_backbone": 1e-05
552
+ },
553
+ "output_dir": "C:\\Users\\Usuario\\Documents\\Workspace\\LeRobot_\\demos\\06_imitation_learning\\outputs\\sweep_super",
554
+ "job_name": "act_so101_sweep_super",
555
+ "resume": false,
556
+ "seed": 1000,
557
+ "num_workers": 4,
558
+ "batch_size": 8,
559
+ "steps": 30000,
560
+ "eval_freq": 20000,
561
+ "log_freq": 100,
562
+ "tolerance_s": 0.0001,
563
+ "save_checkpoint": true,
564
+ "save_freq": 6000,
565
+ "use_policy_training_preset": true,
566
+ "optimizer": {
567
+ "type": "adamw",
568
+ "lr": 1e-05,
569
+ "weight_decay": 0.0001,
570
+ "grad_clip_norm": 10.0,
571
+ "betas": [
572
+ 0.9,
573
+ 0.999
574
+ ],
575
+ "eps": 1e-08
576
+ },
577
+ "scheduler": null,
578
+ "eval": {
579
+ "n_episodes": 50,
580
+ "batch_size": 50,
581
+ "use_async_envs": false
582
+ },
583
+ "wandb": {
584
+ "enable": false,
585
+ "disable_artifact": false,
586
+ "project": "lerobot",
587
+ "entity": null,
588
+ "notes": null,
589
+ "run_id": null,
590
+ "mode": null
591
+ },
592
+ "peft": null,
593
+ "use_rabc": false,
594
+ "rabc_progress_path": null,
595
+ "rabc_kappa": 0.01,
596
+ "rabc_epsilon": 1e-06,
597
+ "rabc_head_mode": "sparse",
598
+ "rename_map": {},
599
+ "checkpoint_path": null
600
+ }
pick_v3_super/checkpoints/030000/dataset_features.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "action": {
3
+ "dtype": "float32",
4
+ "names": [
5
+ "shoulder_pan.pos",
6
+ "shoulder_lift.pos",
7
+ "elbow_flex.pos",
8
+ "wrist_flex.pos",
9
+ "wrist_roll.pos",
10
+ "gripper.pos"
11
+ ],
12
+ "shape": [
13
+ 6
14
+ ]
15
+ },
16
+ "observation.images.front": {
17
+ "dtype": "video",
18
+ "shape": [
19
+ 480,
20
+ 640,
21
+ 3
22
+ ],
23
+ "names": [
24
+ "height",
25
+ "width",
26
+ "channels"
27
+ ],
28
+ "info": {
29
+ "is_depth_map": false,
30
+ "video.height": 480,
31
+ "video.width": 640,
32
+ "video.codec": "h264",
33
+ "video.pix_fmt": "yuv420p",
34
+ "video.fps": 30,
35
+ "video.channels": 3,
36
+ "has_audio": false,
37
+ "video.g": 2,
38
+ "video.crf": 30,
39
+ "video.preset": "p1",
40
+ "video.fast_decode": 0,
41
+ "video.video_backend": "pyav",
42
+ "video.extra_options": {}
43
+ }
44
+ },
45
+ "observation.images.wrist": {
46
+ "dtype": "video",
47
+ "shape": [
48
+ 480,
49
+ 640,
50
+ 3
51
+ ],
52
+ "names": [
53
+ "height",
54
+ "width",
55
+ "channels"
56
+ ],
57
+ "info": {
58
+ "is_depth_map": false,
59
+ "video.height": 480,
60
+ "video.width": 640,
61
+ "video.codec": "h264",
62
+ "video.pix_fmt": "yuv420p",
63
+ "video.fps": 30,
64
+ "video.channels": 3,
65
+ "has_audio": false,
66
+ "video.g": 2,
67
+ "video.crf": 30,
68
+ "video.preset": "p1",
69
+ "video.fast_decode": 0,
70
+ "video.video_backend": "pyav",
71
+ "video.extra_options": {}
72
+ }
73
+ },
74
+ "observation.state": {
75
+ "dtype": "float32",
76
+ "names": [
77
+ "shoulder_pan.pos",
78
+ "shoulder_lift.pos",
79
+ "elbow_flex.pos",
80
+ "wrist_flex.pos",
81
+ "wrist_roll.pos",
82
+ "gripper.pos"
83
+ ],
84
+ "shape": [
85
+ 6
86
+ ]
87
+ }
88
+ }
pick_v3_super/checkpoints/030000/model_card_original.md ADDED
@@ -0,0 +1,49 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ library_name: lerobot
4
+ pipeline_tag: robotics
5
+ tags:
6
+ - lerobot
7
+ - act
8
+ - so101
9
+ - pick
10
+ ---
11
+
12
+ # act-pick-v3-super-30000
13
+
14
+ Pick policy trained on the "super" mix of the third round (v3): every recording that passes the quality filter. Checkpoint at 30,000 steps. This is not the best checkpoint of its run (the best one is at 6,000 steps); it is published for completeness.
15
+
16
+ ACT policy for the SO-101 "pick a piece from the crate and hold it up to the camera" task. Inputs: front camera, wrist camera and the six joint positions; output: a chunk of 90 future joint targets at 30 fps.
17
+
18
+ ## Training data
19
+
20
+ 398 training episodes: all 66 pick-cube sessions, 30 fps and 10 fps (up-sampled to 30 fps), any number of pieces in the crate; 5 to 15 s long, start gap at most 20 degrees, at most 3 s idle at either end, gripper opens at least once, known bad takes removed.
21
+
22
+ Recipe: LeRobot `train()`, batch 8, chunk 90, 30 action steps per replan, default image augmentation, seed 1000.
23
+
24
+ ## Offline check on the held-out episodes
25
+
26
+ Held out: 16 episodes, 12 from `so101-pick-cube_20261024_181939` (one piece) and 4 from `so101-pick-cube_20261024_144537` (two pieces). Mean absolute joint error over the predicted chunk (lower is better) against a "stay at the current joint positions" baseline of 20.71:
27
+
28
+ | checkpoint of this run | error | below baseline |
29
+ |---|---|---|
30
+ | 6,000 steps | 15.16 | 27% |
31
+ | 12,000 steps | 16.08 | 22% |
32
+ | 18,000 steps | 16.04 | 23% |
33
+ | 24,000 steps | 16.36 | 21% |
34
+ | **30,000 steps (this one)** | **16.27** | 21% |
35
+
36
+ All runs of this round, best checkpoint of each:
37
+
38
+ | run | training episodes | best checkpoint | error | below stay-still | repo |
39
+ |---|---|---|---|---|---|
40
+ | super (all quality-passing data) | 398 | 6k | 15.16 | 27% | act-pick-v3-6000 |
41
+ | strict (6-12 s, tight quality) | 205 | 2.4k | 15.14 | 27% | act-pick-v3-strict-2400 |
42
+ | v3 recipe on 1-2 piece episodes | 331 | 18k | 15.35 | 26% | act-pick-v3-1-2-pieces-18000 |
43
+ | v2 (earlier) | 278 | 6k | 15.36 | 26% | act-pick-v2-6000 |
44
+ | chunk 60 (own baseline 16.23) | 331 | 4.8k | 11.94 | 26% | act-pick-v3-chunk60-4800 |
45
+ | one piece only | 292 | 9.6k | 15.47 | 25% | act-pick-v3-1piece-9600 |
46
+ | recent sessions only | 100 | 12k | 16.87 | 19% | act-pick-v3-recent-12000 |
47
+ | first run | 165 | 6k | 16.87 | 19% | act-pick-* |
48
+
49
+ The best checkpoints of the main runs are within noise of each other, and every run got worse after about one pass over the data. The offline error cannot say which model grasps best; none of them has been tested on the robot.
pick_v3_super/checkpoints/030000/pretrained_model/config.json ADDED
@@ -0,0 +1,71 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "type": "act",
3
+ "n_obs_steps": 1,
4
+ "input_features": {
5
+ "observation.state": {
6
+ "type": "STATE",
7
+ "shape": [
8
+ 6
9
+ ]
10
+ },
11
+ "observation.images.front": {
12
+ "type": "VISUAL",
13
+ "shape": [
14
+ 3,
15
+ 480,
16
+ 640
17
+ ]
18
+ },
19
+ "observation.images.wrist": {
20
+ "type": "VISUAL",
21
+ "shape": [
22
+ 3,
23
+ 480,
24
+ 640
25
+ ]
26
+ }
27
+ },
28
+ "output_features": {
29
+ "action": {
30
+ "type": "ACTION",
31
+ "shape": [
32
+ 6
33
+ ]
34
+ }
35
+ },
36
+ "device": "cuda",
37
+ "use_amp": false,
38
+ "use_peft": false,
39
+ "push_to_hub": false,
40
+ "repo_id": null,
41
+ "private": null,
42
+ "tags": null,
43
+ "license": null,
44
+ "pretrained_path": null,
45
+ "chunk_size": 90,
46
+ "n_action_steps": 30,
47
+ "normalization_mapping": {
48
+ "VISUAL": "MEAN_STD",
49
+ "STATE": "MEAN_STD",
50
+ "ACTION": "MEAN_STD"
51
+ },
52
+ "vision_backbone": "resnet18",
53
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
54
+ "replace_final_stride_with_dilation": false,
55
+ "pre_norm": false,
56
+ "dim_model": 512,
57
+ "n_heads": 8,
58
+ "dim_feedforward": 3200,
59
+ "feedforward_activation": "relu",
60
+ "n_encoder_layers": 4,
61
+ "n_decoder_layers": 1,
62
+ "use_vae": true,
63
+ "latent_dim": 32,
64
+ "n_vae_encoder_layers": 4,
65
+ "temporal_ensemble_coeff": null,
66
+ "dropout": 0.1,
67
+ "kl_weight": 10.0,
68
+ "optimizer_lr": 1e-05,
69
+ "optimizer_weight_decay": 0.0001,
70
+ "optimizer_lr_backbone": 1e-05
71
+ }
pick_v3_super/checkpoints/030000/pretrained_model/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c6e42ba0b7ee0b36fcc9e77895b0d7156ce2c297d64be204a3e4096f2b106ba
3
+ size 206658768
pick_v3_super/checkpoints/030000/pretrained_model/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
+ 6
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
+ }
pick_v3_super/checkpoints/030000/pretrained_model/policy_postprocessor_step_0_unnormalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931ed929c95ef00f6e334882511164d4c9d89f16bfb31eeb212ddecf99897ed7
3
+ size 7552
pick_v3_super/checkpoints/030000/pretrained_model/policy_preprocessor.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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.state": {
27
+ "type": "STATE",
28
+ "shape": [
29
+ 6
30
+ ]
31
+ },
32
+ "observation.images.front": {
33
+ "type": "VISUAL",
34
+ "shape": [
35
+ 3,
36
+ 480,
37
+ 640
38
+ ]
39
+ },
40
+ "observation.images.wrist": {
41
+ "type": "VISUAL",
42
+ "shape": [
43
+ 3,
44
+ 480,
45
+ 640
46
+ ]
47
+ },
48
+ "action": {
49
+ "type": "ACTION",
50
+ "shape": [
51
+ 6
52
+ ]
53
+ }
54
+ },
55
+ "norm_map": {
56
+ "VISUAL": "MEAN_STD",
57
+ "STATE": "MEAN_STD",
58
+ "ACTION": "MEAN_STD"
59
+ }
60
+ },
61
+ "state_file": "policy_preprocessor_step_3_normalizer_processor.safetensors"
62
+ }
63
+ ]
64
+ }
pick_v3_super/checkpoints/030000/pretrained_model/policy_preprocessor_step_3_normalizer_processor.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931ed929c95ef00f6e334882511164d4c9d89f16bfb31eeb212ddecf99897ed7
3
+ size 7552
pick_v3_super/checkpoints/030000/pretrained_model/train_config.json ADDED
@@ -0,0 +1,600 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "dataset": {
3
+ "repo_id": "local/so101_act",
4
+ "root": "C:\\Users\\Usuario\\Documents\\Workspace\\LeRobot_\\demos\\06_imitation_learning\\data\\real_so101\\act_pick_super",
5
+ "episodes": [
6
+ 4,
7
+ 5,
8
+ 6,
9
+ 7,
10
+ 8,
11
+ 9,
12
+ 10,
13
+ 12,
14
+ 15,
15
+ 17,
16
+ 18,
17
+ 19,
18
+ 20,
19
+ 21,
20
+ 23,
21
+ 24,
22
+ 25,
23
+ 27,
24
+ 28,
25
+ 29,
26
+ 30,
27
+ 31,
28
+ 32,
29
+ 33,
30
+ 34,
31
+ 35,
32
+ 36,
33
+ 37,
34
+ 38,
35
+ 39,
36
+ 41,
37
+ 42,
38
+ 43,
39
+ 44,
40
+ 45,
41
+ 46,
42
+ 47,
43
+ 49,
44
+ 50,
45
+ 51,
46
+ 52,
47
+ 53,
48
+ 54,
49
+ 55,
50
+ 56,
51
+ 57,
52
+ 58,
53
+ 59,
54
+ 60,
55
+ 63,
56
+ 64,
57
+ 65,
58
+ 66,
59
+ 67,
60
+ 68,
61
+ 69,
62
+ 70,
63
+ 71,
64
+ 72,
65
+ 73,
66
+ 74,
67
+ 76,
68
+ 77,
69
+ 78,
70
+ 79,
71
+ 84,
72
+ 86,
73
+ 87,
74
+ 88,
75
+ 89,
76
+ 90,
77
+ 91,
78
+ 92,
79
+ 93,
80
+ 94,
81
+ 96,
82
+ 97,
83
+ 98,
84
+ 100,
85
+ 101,
86
+ 102,
87
+ 103,
88
+ 104,
89
+ 105,
90
+ 106,
91
+ 107,
92
+ 109,
93
+ 110,
94
+ 111,
95
+ 112,
96
+ 113,
97
+ 114,
98
+ 115,
99
+ 116,
100
+ 117,
101
+ 118,
102
+ 119,
103
+ 120,
104
+ 121,
105
+ 122,
106
+ 124,
107
+ 125,
108
+ 126,
109
+ 127,
110
+ 128,
111
+ 129,
112
+ 130,
113
+ 131,
114
+ 132,
115
+ 133,
116
+ 134,
117
+ 135,
118
+ 136,
119
+ 137,
120
+ 138,
121
+ 139,
122
+ 140,
123
+ 141,
124
+ 142,
125
+ 143,
126
+ 144,
127
+ 145,
128
+ 146,
129
+ 147,
130
+ 148,
131
+ 149,
132
+ 150,
133
+ 151,
134
+ 152,
135
+ 153,
136
+ 154,
137
+ 155,
138
+ 156,
139
+ 157,
140
+ 158,
141
+ 159,
142
+ 160,
143
+ 161,
144
+ 162,
145
+ 163,
146
+ 164,
147
+ 165,
148
+ 166,
149
+ 167,
150
+ 168,
151
+ 169,
152
+ 170,
153
+ 171,
154
+ 172,
155
+ 173,
156
+ 174,
157
+ 175,
158
+ 176,
159
+ 177,
160
+ 178,
161
+ 179,
162
+ 180,
163
+ 181,
164
+ 182,
165
+ 183,
166
+ 184,
167
+ 185,
168
+ 186,
169
+ 187,
170
+ 188,
171
+ 189,
172
+ 190,
173
+ 191,
174
+ 192,
175
+ 193,
176
+ 194,
177
+ 195,
178
+ 196,
179
+ 197,
180
+ 198,
181
+ 199,
182
+ 200,
183
+ 201,
184
+ 202,
185
+ 203,
186
+ 204,
187
+ 205,
188
+ 206,
189
+ 207,
190
+ 208,
191
+ 209,
192
+ 210,
193
+ 211,
194
+ 212,
195
+ 213,
196
+ 214,
197
+ 215,
198
+ 216,
199
+ 217,
200
+ 218,
201
+ 219,
202
+ 220,
203
+ 221,
204
+ 222,
205
+ 223,
206
+ 224,
207
+ 225,
208
+ 226,
209
+ 227,
210
+ 228,
211
+ 229,
212
+ 230,
213
+ 231,
214
+ 232,
215
+ 233,
216
+ 235,
217
+ 236,
218
+ 237,
219
+ 238,
220
+ 239,
221
+ 240,
222
+ 241,
223
+ 242,
224
+ 243,
225
+ 244,
226
+ 245,
227
+ 246,
228
+ 247,
229
+ 248,
230
+ 249,
231
+ 250,
232
+ 251,
233
+ 252,
234
+ 253,
235
+ 255,
236
+ 256,
237
+ 257,
238
+ 258,
239
+ 259,
240
+ 260,
241
+ 261,
242
+ 262,
243
+ 263,
244
+ 264,
245
+ 265,
246
+ 266,
247
+ 267,
248
+ 268,
249
+ 269,
250
+ 270,
251
+ 271,
252
+ 272,
253
+ 273,
254
+ 274,
255
+ 275,
256
+ 276,
257
+ 277,
258
+ 278,
259
+ 279,
260
+ 280,
261
+ 281,
262
+ 282,
263
+ 283,
264
+ 284,
265
+ 285,
266
+ 286,
267
+ 287,
268
+ 288,
269
+ 289,
270
+ 290,
271
+ 291,
272
+ 292,
273
+ 293,
274
+ 294,
275
+ 295,
276
+ 296,
277
+ 297,
278
+ 298,
279
+ 299,
280
+ 300,
281
+ 301,
282
+ 302,
283
+ 303,
284
+ 304,
285
+ 305,
286
+ 306,
287
+ 307,
288
+ 308,
289
+ 309,
290
+ 310,
291
+ 311,
292
+ 312,
293
+ 313,
294
+ 314,
295
+ 315,
296
+ 316,
297
+ 317,
298
+ 318,
299
+ 319,
300
+ 320,
301
+ 325,
302
+ 326,
303
+ 327,
304
+ 329,
305
+ 330,
306
+ 331,
307
+ 332,
308
+ 333,
309
+ 334,
310
+ 335,
311
+ 336,
312
+ 337,
313
+ 338,
314
+ 339,
315
+ 340,
316
+ 341,
317
+ 342,
318
+ 343,
319
+ 344,
320
+ 345,
321
+ 346,
322
+ 347,
323
+ 348,
324
+ 349,
325
+ 351,
326
+ 352,
327
+ 353,
328
+ 354,
329
+ 355,
330
+ 356,
331
+ 357,
332
+ 358,
333
+ 359,
334
+ 360,
335
+ 361,
336
+ 364,
337
+ 365,
338
+ 366,
339
+ 367,
340
+ 368,
341
+ 369,
342
+ 370,
343
+ 371,
344
+ 372,
345
+ 373,
346
+ 374,
347
+ 375,
348
+ 376,
349
+ 377,
350
+ 378,
351
+ 391,
352
+ 392,
353
+ 393,
354
+ 394,
355
+ 395,
356
+ 396,
357
+ 397,
358
+ 398,
359
+ 399,
360
+ 400,
361
+ 401,
362
+ 402,
363
+ 403,
364
+ 404,
365
+ 405,
366
+ 406,
367
+ 407,
368
+ 408,
369
+ 409,
370
+ 410,
371
+ 411,
372
+ 412,
373
+ 413,
374
+ 414,
375
+ 415,
376
+ 416,
377
+ 417,
378
+ 419,
379
+ 420,
380
+ 421,
381
+ 422,
382
+ 423,
383
+ 424,
384
+ 425,
385
+ 426,
386
+ 427,
387
+ 428,
388
+ 429,
389
+ 430,
390
+ 431,
391
+ 432,
392
+ 433,
393
+ 434,
394
+ 435,
395
+ 436,
396
+ 438,
397
+ 439,
398
+ 440,
399
+ 441,
400
+ 442,
401
+ 443,
402
+ 444,
403
+ 445
404
+ ],
405
+ "image_transforms": {
406
+ "enable": true,
407
+ "max_num_transforms": 3,
408
+ "random_order": false,
409
+ "tfs": {
410
+ "brightness": {
411
+ "weight": 1.0,
412
+ "type": "ColorJitter",
413
+ "kwargs": {
414
+ "brightness": [
415
+ 0.8,
416
+ 1.2
417
+ ]
418
+ }
419
+ },
420
+ "contrast": {
421
+ "weight": 1.0,
422
+ "type": "ColorJitter",
423
+ "kwargs": {
424
+ "contrast": [
425
+ 0.8,
426
+ 1.2
427
+ ]
428
+ }
429
+ },
430
+ "saturation": {
431
+ "weight": 1.0,
432
+ "type": "ColorJitter",
433
+ "kwargs": {
434
+ "saturation": [
435
+ 0.5,
436
+ 1.5
437
+ ]
438
+ }
439
+ },
440
+ "hue": {
441
+ "weight": 1.0,
442
+ "type": "ColorJitter",
443
+ "kwargs": {
444
+ "hue": [
445
+ -0.05,
446
+ 0.05
447
+ ]
448
+ }
449
+ },
450
+ "sharpness": {
451
+ "weight": 1.0,
452
+ "type": "SharpnessJitter",
453
+ "kwargs": {
454
+ "sharpness": [
455
+ 0.5,
456
+ 1.5
457
+ ]
458
+ }
459
+ },
460
+ "affine": {
461
+ "weight": 1.0,
462
+ "type": "RandomAffine",
463
+ "kwargs": {
464
+ "degrees": [
465
+ -5.0,
466
+ 5.0
467
+ ],
468
+ "translate": [
469
+ 0.05,
470
+ 0.05
471
+ ]
472
+ }
473
+ }
474
+ }
475
+ },
476
+ "revision": null,
477
+ "use_imagenet_stats": true,
478
+ "video_backend": "pyav",
479
+ "streaming": false
480
+ },
481
+ "env": null,
482
+ "policy": {
483
+ "type": "act",
484
+ "n_obs_steps": 1,
485
+ "input_features": {
486
+ "observation.state": {
487
+ "type": "STATE",
488
+ "shape": [
489
+ 6
490
+ ]
491
+ },
492
+ "observation.images.front": {
493
+ "type": "VISUAL",
494
+ "shape": [
495
+ 3,
496
+ 480,
497
+ 640
498
+ ]
499
+ },
500
+ "observation.images.wrist": {
501
+ "type": "VISUAL",
502
+ "shape": [
503
+ 3,
504
+ 480,
505
+ 640
506
+ ]
507
+ }
508
+ },
509
+ "output_features": {
510
+ "action": {
511
+ "type": "ACTION",
512
+ "shape": [
513
+ 6
514
+ ]
515
+ }
516
+ },
517
+ "device": "cuda",
518
+ "use_amp": false,
519
+ "use_peft": false,
520
+ "push_to_hub": false,
521
+ "repo_id": null,
522
+ "private": null,
523
+ "tags": null,
524
+ "license": null,
525
+ "pretrained_path": null,
526
+ "chunk_size": 90,
527
+ "n_action_steps": 30,
528
+ "normalization_mapping": {
529
+ "VISUAL": "MEAN_STD",
530
+ "STATE": "MEAN_STD",
531
+ "ACTION": "MEAN_STD"
532
+ },
533
+ "vision_backbone": "resnet18",
534
+ "pretrained_backbone_weights": "ResNet18_Weights.IMAGENET1K_V1",
535
+ "replace_final_stride_with_dilation": false,
536
+ "pre_norm": false,
537
+ "dim_model": 512,
538
+ "n_heads": 8,
539
+ "dim_feedforward": 3200,
540
+ "feedforward_activation": "relu",
541
+ "n_encoder_layers": 4,
542
+ "n_decoder_layers": 1,
543
+ "use_vae": true,
544
+ "latent_dim": 32,
545
+ "n_vae_encoder_layers": 4,
546
+ "temporal_ensemble_coeff": null,
547
+ "dropout": 0.1,
548
+ "kl_weight": 10.0,
549
+ "optimizer_lr": 1e-05,
550
+ "optimizer_weight_decay": 0.0001,
551
+ "optimizer_lr_backbone": 1e-05
552
+ },
553
+ "output_dir": "C:\\Users\\Usuario\\Documents\\Workspace\\LeRobot_\\demos\\06_imitation_learning\\outputs\\sweep_super",
554
+ "job_name": "act_so101_sweep_super",
555
+ "resume": false,
556
+ "seed": 1000,
557
+ "num_workers": 4,
558
+ "batch_size": 8,
559
+ "steps": 30000,
560
+ "eval_freq": 20000,
561
+ "log_freq": 100,
562
+ "tolerance_s": 0.0001,
563
+ "save_checkpoint": true,
564
+ "save_freq": 6000,
565
+ "use_policy_training_preset": true,
566
+ "optimizer": {
567
+ "type": "adamw",
568
+ "lr": 1e-05,
569
+ "weight_decay": 0.0001,
570
+ "grad_clip_norm": 10.0,
571
+ "betas": [
572
+ 0.9,
573
+ 0.999
574
+ ],
575
+ "eps": 1e-08
576
+ },
577
+ "scheduler": null,
578
+ "eval": {
579
+ "n_episodes": 50,
580
+ "batch_size": 50,
581
+ "use_async_envs": false
582
+ },
583
+ "wandb": {
584
+ "enable": false,
585
+ "disable_artifact": false,
586
+ "project": "lerobot",
587
+ "entity": null,
588
+ "notes": null,
589
+ "run_id": null,
590
+ "mode": null
591
+ },
592
+ "peft": null,
593
+ "use_rabc": false,
594
+ "rabc_progress_path": null,
595
+ "rabc_kappa": 0.01,
596
+ "rabc_epsilon": 1e-06,
597
+ "rabc_head_mode": "sparse",
598
+ "rename_map": {},
599
+ "checkpoint_path": null
600
+ }