ds1x commited on
Commit
224a65c
·
verified ·
1 Parent(s): c829c86

Publish complete RSCI 40000 checkpoint: newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045

Browse files
Files changed (15) hide show
  1. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/argv.txt +64 -0
  2. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoint-manifest.json +99 -0
  3. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/model.safetensors +3 -0
  4. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/optimizer.bin +3 -0
  5. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_0.pkl +3 -0
  6. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_1.pkl +3 -0
  7. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_2.pkl +3 -0
  8. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_3.pkl +3 -0
  9. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_4.pkl +3 -0
  10. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_5.pkl +3 -0
  11. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_6.pkl +3 -0
  12. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_7.pkl +3 -0
  13. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/scheduler.bin +3 -0
  14. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/launch_contract.json +417 -0
  15. newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/run_config.json +434 -0
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/argv.txt ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ /root/workspace/Haoxuan/code/Psi0/RSCI/feture/ARDY_PSCI/scripts/train.py
2
+ finetune_simple_psi0_config
3
+ --train.name=finetune
4
+ --train.data_parallel=ddp
5
+ --train.mixed_precision=bf16
6
+ --train.gradient_accumulation_steps=1
7
+ --train.learning_rate=1e-4
8
+ --train.max_training_steps=40000
9
+ --train.warmup_ratio=None
10
+ --train.warmup_steps=1000
11
+ --train.checkpointing_steps=1000
12
+ --train.checkpoint-anchor-steps=10000
13
+ --train.max_checkpoints_to_keep=4
14
+ --train.validation_steps=500
15
+ --train.val_num_batches=20
16
+ --train.max_grad_norm=1.0
17
+ --train.lr_scheduler_type=cosine
18
+ --train.lr_scheduler_kwargs.weight_decay=1e-6
19
+ --train.lr_scheduler_kwargs.betas 0.95 0.999
20
+ --data.transform.repack.pad-action-dim=36
21
+ --data.transform.repack.pad-state-dim=36
22
+ --data.transform.field.stat-path=meta/stats_psi0.json
23
+ --data.transform.field.stat-action-key=action
24
+ --data.transform.field.stat-state-key=states
25
+ --data.transform.field.action_norm_type=bounds
26
+ --data.transform.field.no-use-norm-mask
27
+ --data.transform.field.normalize-state
28
+ --data.transform.field.pad-action-dim=36
29
+ --data.transform.field.pad-state-dim=36
30
+ --data.transform.model.img-aug
31
+ --data.transform.model.resize.size 180 320
32
+ --data.transform.model.center_crop.size 180 320
33
+ --model.noise-scheduler=flow
34
+ --model.train-diffusion-steps=1000
35
+ --model.n_conditions=0
36
+ --model.action-chunk-size=30
37
+ --model.action-dim=36
38
+ --model.action-exec-horizon=30
39
+ --model.observation-horizon=1
40
+ --model.odim=36
41
+ --model.view_feature_dim=2048
42
+ --model.no-use_film
43
+ --model.no-combined_temb
44
+ --model.rtc
45
+ --model.max-delay=8
46
+ --model.psi0-ardy-aux-weight=0.1
47
+ --exp=RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1
48
+ --seed=292285
49
+ --train.output-dir=/root/workspace/Haoxuan/data/Psi0/RSCI/.runs/G1WholebodyLocomotionPickBetweenTablesTeleop-v0/VGQ-AE/frozen/seed-292285/newh264-speed-v1
50
+ --train.train_batch_size=16
51
+ --train.num-workers=8
52
+ --log.report_to=None
53
+ --train.val-num-workers=8
54
+ --train.prefetch-factor=2
55
+ --data.root_dir=/root/workspace/Haoxuan/data/Psi0/data/simple-between-tables-new-training-h264
56
+ --data.train-repo-ids=G1WholebodyLocomotionPickBetweenTablesTeleop-v0
57
+ --model.model_name_or_path=/root/workspace/Haoxuan/data/Psi0/cache/checkpoints/psi0/pre.fast.1by1.2601091803.ckpt.ego200k.he30k
58
+ --model.pretrained-action-header-path=/root/workspace/Haoxuan/data/Psi0/cache/checkpoints/psi0/postpre.1by1.pad36.2601131206.ckpt.he30k
59
+ --model.rsci-variant=VGQ-AE
60
+ --model.rsci-vlm-mode=frozen
61
+ --data.v5-ardy-sidecar-root=/root/workspace/Haoxuan/data/Psi0/data/simple-between-tables-new-training-ardy-h264/G1WholebodyLocomotionPickBetweenTablesTeleop-v0
62
+ --train.optimizer-fused
63
+ --train.pin-memory
64
+ --train.non-blocking-transfer
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoint-manifest.json ADDED
@@ -0,0 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema": 1,
3
+ "source_run": "/root/workspace/Haoxuan/data/Psi0/RSCI/.runs/G1WholebodyLocomotionPickBetweenTablesTeleop-v0/VGQ-AE/frozen/seed-292285/newh264-speed-v1/finetune/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045",
4
+ "task": "G1WholebodyLocomotionPickBetweenTablesTeleop-v0",
5
+ "model": "VGQ-AE",
6
+ "vlm": "frozen",
7
+ "step": 40000,
8
+ "world_size": 8,
9
+ "prefix": "newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045",
10
+ "total_bytes": 10289936699,
11
+ "files": {
12
+ "checkpoints/ckpt_40000/model.safetensors": {
13
+ "bytes": 6292351720,
14
+ "sha256": "747fb503d84b13ecbc27f3748b5829afa77a6068621de30470f2f7eac110d0e1",
15
+ "git_blob_id": null,
16
+ "mtime_ns": 1792248090373475560
17
+ },
18
+ "checkpoints/ckpt_40000/optimizer.bin": {
19
+ "bytes": 3997422027,
20
+ "sha256": "f23852547108e7c7c7ab17298f3872d9182cde82e7b2e8c67e5ec33b14636369",
21
+ "git_blob_id": null,
22
+ "mtime_ns": 1792248097145479130
23
+ },
24
+ "checkpoints/ckpt_40000/random_states_0.pkl": {
25
+ "bytes": 16577,
26
+ "sha256": "fb1f9c1b86e8f418f19dfef4411ccd3489caee55877a64da78e21df20f5f53d5",
27
+ "git_blob_id": "e0fe035873f308416fa51ff47754c16e95306b94",
28
+ "mtime_ns": 1792248097152479134
29
+ },
30
+ "checkpoints/ckpt_40000/random_states_1.pkl": {
31
+ "bytes": 16513,
32
+ "sha256": "c28c015cb251af89140d520bf7b787e2c1a96f31c7093fb8980b77717c6d9a7c",
33
+ "git_blob_id": "476813204bd4a708ca49d422a49b8ce6f553a145",
34
+ "mtime_ns": 1792248078046469061
35
+ },
36
+ "checkpoints/ckpt_40000/random_states_2.pkl": {
37
+ "bytes": 16513,
38
+ "sha256": "af10a5f1fe7547b399123c0ebdeadac3fca26903049f8381abc4bab22823a96d",
39
+ "git_blob_id": "2cd3622727c1aed47b606773cedd1f6b710fda35",
40
+ "mtime_ns": 1792248078046469061
41
+ },
42
+ "checkpoints/ckpt_40000/random_states_3.pkl": {
43
+ "bytes": 16513,
44
+ "sha256": "eb541a74149cc3b731c99f091de596a417ae64b26672dfbdbb64cc72dccc1e1e",
45
+ "git_blob_id": "1d615513b69090274932173b2b1f1942816750ea",
46
+ "mtime_ns": 1792248078048469062
47
+ },
48
+ "checkpoints/ckpt_40000/random_states_4.pkl": {
49
+ "bytes": 16513,
50
+ "sha256": "8d957016d0a1b8963b9fd485a7604893b99529e74b9a5261de47ba8c349524d8",
51
+ "git_blob_id": "aac7cf9315e962ae4c40fc7258d51300dcecc8dc",
52
+ "mtime_ns": 1792248078046469061
53
+ },
54
+ "checkpoints/ckpt_40000/random_states_5.pkl": {
55
+ "bytes": 16513,
56
+ "sha256": "a609e3e32ea661456d884ab727d9b1b042c22d815851a09fc4677d87ba7b14c8",
57
+ "git_blob_id": "d2ea12a2b234222495a337166132ce57ba331269",
58
+ "mtime_ns": 1792248078046469061
59
+ },
60
+ "checkpoints/ckpt_40000/random_states_6.pkl": {
61
+ "bytes": 16513,
62
+ "sha256": "694b0a5c6a6e2ecf4824196058ac6e5cd2d074d93858094d8551ef59d4783da7",
63
+ "git_blob_id": "5ea01d2e63bb825e26bb684bdca1075d28e3abd5",
64
+ "mtime_ns": 1792248078055469066
65
+ },
66
+ "checkpoints/ckpt_40000/random_states_7.pkl": {
67
+ "bytes": 16513,
68
+ "sha256": "58324d7740510cc4a8638dc67cfedfa2748a88c4acfbcafaf907fd751ce59c8b",
69
+ "git_blob_id": "f7d27b4b2763115a7bb9af21118427ac10c995fb",
70
+ "mtime_ns": 1792248078046469061
71
+ },
72
+ "checkpoints/ckpt_40000/scheduler.bin": {
73
+ "bytes": 1401,
74
+ "sha256": "333a511cd62b9f69fddb60ba012421d8c02880f26090feecb0440de01e609e64",
75
+ "git_blob_id": "579b8a5952a2274f9241e922390e3fef6f18e8b8",
76
+ "mtime_ns": 1792248097147479131
77
+ },
78
+ "run_config.json": {
79
+ "bytes": 15050,
80
+ "sha256": "5618d84f73c9f13554bef04f0d3e7b5098f511158270b8c9583f4bffdaa56856",
81
+ "git_blob_id": "144add04e9f98bc27f400170b95361e3dc5bd0bb",
82
+ "mtime_ns": 1792205113129056366
83
+ },
84
+ "argv.txt": {
85
+ "bytes": 2652,
86
+ "sha256": "ac5a64a7408aaba0ad75e06aabd35e68ba397f0acb081cce0bac5ba946c0ee23",
87
+ "git_blob_id": "3b99349efd7c1a5eae452840ca9caf6751515e2d",
88
+ "mtime_ns": 1792205113131056367
89
+ },
90
+ "launch_contract.json": {
91
+ "bytes": 11681,
92
+ "sha256": "56bd301ce949258e41a82322c386d5237fde0f40995ea41ba6d819b5ee077888",
93
+ "git_blob_id": "1fecdc22727268e347efb664ff0196b2e21796b8",
94
+ "mtime_ns": 1792205090716043770
95
+ }
96
+ },
97
+ "launch_source_sha256": "9a6e2b02c7c64f7942c3f5f02c2cd0333b6e46ea12038e0c41cdb6598df98cdb",
98
+ "contents": "full checkpoint including optimizer, scheduler and all rank RNG states; configs included"
99
+ }
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:747fb503d84b13ecbc27f3748b5829afa77a6068621de30470f2f7eac110d0e1
3
+ size 6292351720
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/optimizer.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f23852547108e7c7c7ab17298f3872d9182cde82e7b2e8c67e5ec33b14636369
3
+ size 3997422027
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_0.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb1f9c1b86e8f418f19dfef4411ccd3489caee55877a64da78e21df20f5f53d5
3
+ size 16577
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_1.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c28c015cb251af89140d520bf7b787e2c1a96f31c7093fb8980b77717c6d9a7c
3
+ size 16513
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_2.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af10a5f1fe7547b399123c0ebdeadac3fca26903049f8381abc4bab22823a96d
3
+ size 16513
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_3.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eb541a74149cc3b731c99f091de596a417ae64b26672dfbdbb64cc72dccc1e1e
3
+ size 16513
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_4.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d957016d0a1b8963b9fd485a7604893b99529e74b9a5261de47ba8c349524d8
3
+ size 16513
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_5.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a609e3e32ea661456d884ab727d9b1b042c22d815851a09fc4677d87ba7b14c8
3
+ size 16513
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_6.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:694b0a5c6a6e2ecf4824196058ac6e5cd2d074d93858094d8551ef59d4783da7
3
+ size 16513
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/random_states_7.pkl ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58324d7740510cc4a8638dc67cfedfa2748a88c4acfbcafaf907fd751ce59c8b
3
+ size 16513
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/checkpoints/ckpt_40000/scheduler.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:333a511cd62b9f69fddb60ba012421d8c02880f26090feecb0440de01e609e64
3
+ size 1401
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/launch_contract.json ADDED
@@ -0,0 +1,417 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema": 1,
3
+ "model": {
4
+ "resnet_store_path": null,
5
+ "pretrained_action_header_path": "/root/workspace/Haoxuan/data/Psi0/cache/checkpoints/psi0/postpre.1by1.pad36.2601131206.ckpt.he30k",
6
+ "rtc": true,
7
+ "max_delay": 8,
8
+ "action_dim": 36,
9
+ "action_chunk_size": 30,
10
+ "action_exec_horizon": 30,
11
+ "observation_horizon": 1,
12
+ "img_chunk": 1,
13
+ "n_cams": 1,
14
+ "use_obs": "add_token",
15
+ "rsci_variant": "VGQ-AE",
16
+ "rsci_vlm_mode": "frozen",
17
+ "rsci_lora_rank": 8,
18
+ "rsci_lora_alpha": 16.0,
19
+ "psi0_ardy_aux": false,
20
+ "psi0_ardy_aux_weight": 0.1,
21
+ "dropout": 0.1,
22
+ "noise_scheduler": "flow",
23
+ "train_diffusion_steps": 1000,
24
+ "eval_diffusion_steps": 10,
25
+ "share_cam_features": false,
26
+ "early_fusion": false,
27
+ "odim": 36,
28
+ "n_conditions": 0,
29
+ "token_fusion": "concat",
30
+ "loss_w": [
31
+ 0.1,
32
+ 0.2,
33
+ 0.1
34
+ ],
35
+ "time_dim": 256,
36
+ "hidden_dim": 1536,
37
+ "num_blocks": 6,
38
+ "dim_feedforward": 2048,
39
+ "nhead": 24,
40
+ "activation": "gelu",
41
+ "view_feature_dim": 2048,
42
+ "use_film": false,
43
+ "combined_temb": false,
44
+ "use_dit": false,
45
+ "use_depth": false,
46
+ "depth_encoder_name": "defm_vit_s14",
47
+ "depth_encoder_path": null,
48
+ "depth_encoder_dim": 384,
49
+ "depth_encoder_input_size": [
50
+ 196,
51
+ 350
52
+ ],
53
+ "depth_encoder_patch_size": 14,
54
+ "freeze_depth_encoder": true,
55
+ "depth_cross_attention_layers": [
56
+ 0,
57
+ 1,
58
+ 2,
59
+ 3,
60
+ 4,
61
+ 5
62
+ ],
63
+ "depth_fusion_mode": "cross_attention",
64
+ "depth_query_count": 16,
65
+ "depth_query_target_count": 8.0,
66
+ "depth_query_score_dim": 384,
67
+ "depth_query_selection": true,
68
+ "depth_query_use_state": true,
69
+ "depth_query_fuse": false,
70
+ "depth_query_align_depth": false,
71
+ "depth_query_independent_rgb": false,
72
+ "depth_query_concat_streams": false,
73
+ "depth_query_cross_attention_fusion": false,
74
+ "depth_rgb_encoder_path": null,
75
+ "depth_rgb_encoder_type": "dinov2",
76
+ "depth_query_pre_orthogonality": false,
77
+ "depth_query_pre_orthogonality_weight": 0.01,
78
+ "depth_query_post_orthogonality": true,
79
+ "depth_query_resampler_max_tokens": 4096,
80
+ "depth_query_gate_init": 1.0,
81
+ "depth_query_sparsity_init": 0.01,
82
+ "depth_query_sparsity_rate": 0.01,
83
+ "depth_query_sparsity_warmup_steps": 1000,
84
+ "depth_query_recovery_after_steps": 100,
85
+ "depth_query_recovery_weight": 0.01,
86
+ "depth_query_recovery_window_steps": 20,
87
+ "depth_query_recovery_sparsity_scale": 0.0,
88
+ "depth_query_alive_margin": 0.1,
89
+ "depth_query_alive_weight": 1.0,
90
+ "depth_query_selection_weight": 0.01,
91
+ "depth_query_orthogonality_weight": 0.01,
92
+ "v5_enabled": false,
93
+ "v5_film": true,
94
+ "v5_pooling": "attention",
95
+ "v5_ardy": true,
96
+ "v5_hand_state": false,
97
+ "v5_object": true,
98
+ "v5_time_encoding": true,
99
+ "v5_temporal_layers": 0,
100
+ "v5_ardy_weight": 0.1,
101
+ "v5_object_weight": 0.1,
102
+ "v5_intra_modal_dedup": false,
103
+ "v5_intra_modal_dedup_weight": 0.01,
104
+ "depth_gate_init": 0.001,
105
+ "depth_module_lr": 0.0001,
106
+ "action_expert_lr": 0.0001,
107
+ "weight_decay": 0.01,
108
+ "model_name_or_path": "/root/workspace/Haoxuan/data/Psi0/cache/checkpoints/psi0/pre.fast.1by1.2601091803.ckpt.ego200k.he30k",
109
+ "vlm_ckpt_step": null,
110
+ "tune_vlm": false,
111
+ "tune_mm_llm": false,
112
+ "tune_mm_vision": false,
113
+ "tune_mm_mlp": false,
114
+ "gradient_checkpointing": true,
115
+ "lang_backbone_lr": 1e-05,
116
+ "mm_projector_lr": 1e-05,
117
+ "vision_tower_lr": 1e-06,
118
+ "optim": "adamw_torch",
119
+ "model_max_length": 4096,
120
+ "data_flatten": true,
121
+ "data_packing": true,
122
+ "max_pixels": 451584,
123
+ "min_pixels": 12544
124
+ },
125
+ "data": {
126
+ "transform": {
127
+ "repack": {
128
+ "dataset_name": "simple",
129
+ "rsci_variant": "VGQ-AE",
130
+ "psi0_ardy_aux": false,
131
+ "v5_ardy": false,
132
+ "v5_object": false,
133
+ "num_past_frames": 0,
134
+ "action_chunk_size": 30,
135
+ "use_depth": false,
136
+ "depth_key": "observation.depth.egocentric",
137
+ "pad_action_dim": 36,
138
+ "pad_state_dim": 36
139
+ },
140
+ "model": {
141
+ "depth_query_independent_rgb": false,
142
+ "resize": {
143
+ "size": [
144
+ 180,
145
+ 320
146
+ ]
147
+ },
148
+ "center_crop": {
149
+ "size": [
150
+ 180,
151
+ 320
152
+ ]
153
+ },
154
+ "color_jitter": {
155
+ "brightness": 0.2,
156
+ "contrast": [
157
+ 0.8,
158
+ 1.2
159
+ ],
160
+ "saturation": [
161
+ 0.8,
162
+ 1.2
163
+ ],
164
+ "hue": 0.05
165
+ },
166
+ "gaussian_noise": {
167
+ "mean": 0.0,
168
+ "std": 3.0,
169
+ "prob_skip": 0.1
170
+ },
171
+ "img_aug": true,
172
+ "adaptive_resize": false,
173
+ "img_sizes": {
174
+ "egodex": [
175
+ 270,
176
+ 480
177
+ ],
178
+ "he": [
179
+ 240,
180
+ 320
181
+ ]
182
+ }
183
+ },
184
+ "field": {
185
+ "stat_path": "meta/stats_psi0.json",
186
+ "action_norm_type": "bounds",
187
+ "stat_action_key": "action",
188
+ "stat_state_key": "states",
189
+ "use_norm_mask": false,
190
+ "action_norm_masks": [
191
+ true,
192
+ true,
193
+ true,
194
+ true,
195
+ true,
196
+ true,
197
+ false
198
+ ],
199
+ "action_min": [
200
+ 0.0,
201
+ 0.0,
202
+ 0.0,
203
+ -1.5,
204
+ -1.5,
205
+ -0.6000000238418579,
206
+ -1.5,
207
+ -0.5,
208
+ -0.699999988079071,
209
+ -0.699999988079071,
210
+ -2.216935230032842e-16,
211
+ -2.216935230032842e-16,
212
+ -4.0845591349633594e-18,
213
+ -2.216935230032842e-16,
214
+ -0.4883034825325012,
215
+ 0.1900009959936142,
216
+ -0.5470856428146362,
217
+ -0.34318920969963074,
218
+ -0.35952919721603394,
219
+ -0.35302427411079407,
220
+ -0.4469815790653229,
221
+ -0.6371198296546936,
222
+ -0.7290776371955872,
223
+ -1.0653810501098633,
224
+ -0.8479154706001282,
225
+ -1.0297260284423828,
226
+ -0.42936205863952637,
227
+ -0.5147944092750549,
228
+ -0.16820405423641205,
229
+ -0.045328833162784576,
230
+ -0.13282617926597595,
231
+ 0.7400000095367432,
232
+ -0.5,
233
+ -0.5,
234
+ -1.0,
235
+ -3.138223648071289
236
+ ],
237
+ "action_max": [
238
+ 0.5,
239
+ 0.699999988079071,
240
+ 0.699999988079071,
241
+ 0.0,
242
+ 0.0,
243
+ 0.0,
244
+ 0.0,
245
+ 6.930528109384597e-19,
246
+ 1.108467615016421e-16,
247
+ 1.108467615016421e-16,
248
+ 1.5,
249
+ 1.5,
250
+ 1.0,
251
+ 1.5,
252
+ 0.2557959258556366,
253
+ 0.35884979367256165,
254
+ 0.5090755820274353,
255
+ 0.19132143259048462,
256
+ 0.21249642968177795,
257
+ 0.3992660641670227,
258
+ 0.4283020794391632,
259
+ 0.1457289606332779,
260
+ -0.1900009959936142,
261
+ 0.6150448322296143,
262
+ 0.35468167066574097,
263
+ 0.8703295588493347,
264
+ 0.7531875371932983,
265
+ 0.971237301826477,
266
+ 0.13985762000083923,
267
+ 0.15686897933483124,
268
+ 0.4661160111427307,
269
+ 0.7400000095367432,
270
+ 0.5,
271
+ 0.5,
272
+ 1.0,
273
+ 3.1414895057678223
274
+ ],
275
+ "state_min": [
276
+ -0.02442001923918724,
277
+ -0.0517612099647522,
278
+ -0.0006534014828503132,
279
+ -0.5095356106758118,
280
+ -1.323034405708313,
281
+ -1.3221508264541626,
282
+ -1.3230019807815552,
283
+ -0.5770347714424133,
284
+ -0.4338151812553406,
285
+ -0.6721642017364502,
286
+ -0.0011892789043486118,
287
+ -7.534810038123396e-07,
288
+ -0.001927333534695208,
289
+ -1.075333216249419e-06,
290
+ -0.43650975823402405,
291
+ 0.15721464157104492,
292
+ -0.5489339232444763,
293
+ -0.26434481143951416,
294
+ -0.3508843183517456,
295
+ -0.23784859478473663,
296
+ -0.4281824827194214,
297
+ -0.5803383588790894,
298
+ -0.6814778447151184,
299
+ -1.0344431400299072,
300
+ -0.7932196259498596,
301
+ -1.0205217599868774,
302
+ -0.3445618450641632,
303
+ -0.5986371040344238,
304
+ -0.1348579376935959,
305
+ -0.0017330688424408436,
306
+ -0.1421850621700287,
307
+ 0.7400000095367432,
308
+ 0.0,
309
+ 0.0,
310
+ 0.0,
311
+ 0.0
312
+ ],
313
+ "state_max": [
314
+ 0.47981399297714233,
315
+ 0.6772664189338684,
316
+ 0.6746510863304138,
317
+ 0.0010172375477850437,
318
+ 0.0007091082516126335,
319
+ 0.001881288131698966,
320
+ 0.0011398319620639086,
321
+ 6.141255539660051e-07,
322
+ 0.3043450713157654,
323
+ 6.343479981296696e-07,
324
+ 0.6933000087738037,
325
+ 1.4612544775009155,
326
+ 1.4651201963424683,
327
+ 1.4609057903289795,
328
+ 0.2809508740901947,
329
+ 0.34028318524360657,
330
+ 0.48582184314727783,
331
+ 0.26476219296455383,
332
+ 0.20825636386871338,
333
+ 0.4566418528556824,
334
+ 0.42864030599594116,
335
+ 0.1656116098165512,
336
+ -0.1549365073442459,
337
+ 0.5154499411582947,
338
+ 0.4242899715900421,
339
+ 0.8548054695129395,
340
+ 0.8040095567703247,
341
+ 0.9811649322509766,
342
+ 0.136736661195755,
343
+ 0.195722296833992,
344
+ 0.45781663060188293,
345
+ 0.7400000095367432,
346
+ 0.0,
347
+ 0.0,
348
+ 0.0,
349
+ 0.0
350
+ ],
351
+ "normalize_state": true,
352
+ "pad_action_dim": 36,
353
+ "pad_state_dim": 36
354
+ }
355
+ },
356
+ "root_dir": "/root/workspace/Haoxuan/data/Psi0/data/simple-between-tables-new-training-h264",
357
+ "v5_ardy_sidecar_root": "/root/workspace/Haoxuan/data/Psi0/data/simple-between-tables-new-training-ardy-h264/G1WholebodyLocomotionPickBetweenTablesTeleop-v0",
358
+ "rgb_frame_cache_root": null,
359
+ "train_repo_ids": [
360
+ "G1WholebodyLocomotionPickBetweenTablesTeleop-v0"
361
+ ],
362
+ "val_repo_ids": [
363
+ "G1WholebodyLocomotionPickBetweenTablesTeleop-v0"
364
+ ]
365
+ },
366
+ "train": {
367
+ "num_workers": 8,
368
+ "overfit_single_batch": false,
369
+ "name": "finetune",
370
+ "hf_token": ".hf_token",
371
+ "lora": false,
372
+ "output_dir": "/root/workspace/Haoxuan/data/Psi0/RSCI/.runs/G1WholebodyLocomotionPickBetweenTablesTeleop-v0/VGQ-AE/frozen/seed-292285/newh264-speed-v1",
373
+ "gradient_accumulation_steps": 1,
374
+ "mixed_precision": "bf16",
375
+ "max_grad_norm": 1.0,
376
+ "optimizer_foreach": null,
377
+ "optimizer_fused": true,
378
+ "pin_memory": true,
379
+ "non_blocking_transfer": true,
380
+ "compile_frozen_encoders": false,
381
+ "val_num_workers": 8,
382
+ "prefetch_factor": 2,
383
+ "timing_enabled": false,
384
+ "reuse_eval_features": false,
385
+ "train_batch_size": 16,
386
+ "val_batch_size": 16,
387
+ "val_num_batches": 20,
388
+ "checkpointing_steps": 1000,
389
+ "checkpoint_anchor_steps": 10000,
390
+ "max_checkpoints_to_keep": 4,
391
+ "validation_steps": 500,
392
+ "learning_rate": 0.0001,
393
+ "lr_scheduler_type": "cosine",
394
+ "lr_scheduler_kwargs": {
395
+ "betas": [
396
+ 0.95,
397
+ 0.999
398
+ ],
399
+ "weight_decay": 1e-06,
400
+ "eps": 1e-08
401
+ },
402
+ "scheduler_specific_kwargs": {},
403
+ "data_parallel": "ddp",
404
+ "sharding_strategy": "full-shard",
405
+ "deepspeed_config": "/root/workspace/Haoxuan/code/Psi0/RSCI/feture/ARDY_PSCI/src/InternVLA/config/deepseeds/zero3.json",
406
+ "enable_gradient_checkpointing": true,
407
+ "enable_mixed_precision_training": true,
408
+ "reduce_in_full_precision": true,
409
+ "max_training_steps": 40000,
410
+ "num_train_epochs": null,
411
+ "warmup_steps": 1000,
412
+ "warmup_ratio": null
413
+ },
414
+ "seed": 292285,
415
+ "world_size": 8,
416
+ "source_sha256": "9a6e2b02c7c64f7942c3f5f02c2cd0333b6e46ea12038e0c41cdb6598df98cdb"
417
+ }
newcheckpoint/Psi0/VGQ-AE/betweentables/frozen/RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1.simple.flow1000.cosine.lr1.0e-04.b128.gpus8.2610171045/run_config.json ADDED
@@ -0,0 +1,434 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "exp": "RSCI-G1WholebodyLocomotionPickBetweenTablesTeleop-v0-VGQ-AE-frozen-newh264-speed-v1",
3
+ "seed": 292285,
4
+ "auto_tag_run": false,
5
+ "eval": false,
6
+ "debug": false,
7
+ "timestamp": "2610171045",
8
+ "log": {
9
+ "logging_dir": "logs",
10
+ "report_to": null,
11
+ "log_freq": 100
12
+ },
13
+ "wandb": {
14
+ "project": "psi",
15
+ "entity": null,
16
+ "group": null,
17
+ "id": null,
18
+ "name": null,
19
+ "resume": "allow"
20
+ },
21
+ "train": {
22
+ "num_workers": 8,
23
+ "overfit_single_batch": false,
24
+ "name": "finetune",
25
+ "resume_from_checkpoint": null,
26
+ "skip_resumed_steps": false,
27
+ "hf_token": ".hf_token",
28
+ "lora": false,
29
+ "output_dir": "/root/workspace/Haoxuan/data/Psi0/RSCI/.runs/G1WholebodyLocomotionPickBetweenTablesTeleop-v0/VGQ-AE/frozen/seed-292285/newh264-speed-v1",
30
+ "gradient_accumulation_steps": 1,
31
+ "mixed_precision": "bf16",
32
+ "max_grad_norm": 1.0,
33
+ "optimizer_foreach": null,
34
+ "optimizer_fused": true,
35
+ "pin_memory": true,
36
+ "non_blocking_transfer": true,
37
+ "compile_frozen_encoders": false,
38
+ "val_num_workers": 8,
39
+ "prefetch_factor": 2,
40
+ "timing_enabled": false,
41
+ "reuse_eval_features": false,
42
+ "train_batch_size": 16,
43
+ "val_batch_size": 16,
44
+ "val_num_batches": 20,
45
+ "checkpointing_steps": 1000,
46
+ "checkpoint_anchor_steps": 10000,
47
+ "max_checkpoints_to_keep": 4,
48
+ "validation_steps": 500,
49
+ "learning_rate": 0.0001,
50
+ "lr_scheduler_type": "cosine",
51
+ "lr_scheduler_kwargs": {
52
+ "betas": [
53
+ 0.95,
54
+ 0.999
55
+ ],
56
+ "weight_decay": 1e-6,
57
+ "eps": 1e-8
58
+ },
59
+ "scheduler_specific_kwargs": {},
60
+ "data_parallel": "ddp",
61
+ "sharding_strategy": "full-shard",
62
+ "deepspeed_config": "/root/workspace/Haoxuan/code/Psi0/RSCI/feture/ARDY_PSCI/src/InternVLA/config/deepseeds/zero3.json",
63
+ "enable_gradient_checkpointing": true,
64
+ "enable_mixed_precision_training": true,
65
+ "reduce_in_full_precision": true,
66
+ "max_training_steps": 40000,
67
+ "num_train_epochs": null,
68
+ "warmup_steps": 1000,
69
+ "warmup_ratio": null
70
+ },
71
+ "data": {
72
+ "transform": {
73
+ "repack": {
74
+ "dataset_name": "simple",
75
+ "rsci_variant": null,
76
+ "psi0_ardy_aux": false,
77
+ "v5_ardy": false,
78
+ "v5_object": false,
79
+ "num_past_frames": 0,
80
+ "action_chunk_size": 30,
81
+ "use_depth": false,
82
+ "depth_key": "observation.depth.egocentric",
83
+ "pad_action_dim": 36,
84
+ "pad_state_dim": 36
85
+ },
86
+ "model": {
87
+ "depth_query_independent_rgb": false,
88
+ "resize": {
89
+ "size": [
90
+ 180,
91
+ 320
92
+ ]
93
+ },
94
+ "center_crop": {
95
+ "size": [
96
+ 180,
97
+ 320
98
+ ]
99
+ },
100
+ "color_jitter": {
101
+ "brightness": 0.2,
102
+ "contrast": [
103
+ 0.8,
104
+ 1.2
105
+ ],
106
+ "saturation": [
107
+ 0.8,
108
+ 1.2
109
+ ],
110
+ "hue": 0.05
111
+ },
112
+ "gaussian_noise": {
113
+ "mean": 0.0,
114
+ "std": 3.0,
115
+ "prob_skip": 0.1
116
+ },
117
+ "img_aug": true,
118
+ "adaptive_resize": false,
119
+ "img_sizes": {
120
+ "egodex": [
121
+ 270,
122
+ 480
123
+ ],
124
+ "he": [
125
+ 240,
126
+ 320
127
+ ]
128
+ }
129
+ },
130
+ "field": {
131
+ "stat_path": "meta/stats_psi0.json",
132
+ "action_norm_type": "bounds",
133
+ "stat_action_key": "action",
134
+ "stat_state_key": "states",
135
+ "use_norm_mask": false,
136
+ "action_norm_masks": [
137
+ true,
138
+ true,
139
+ true,
140
+ true,
141
+ true,
142
+ true,
143
+ false
144
+ ],
145
+ "action_min": [
146
+ 0.0,
147
+ 0.0,
148
+ 0.0,
149
+ -1.5,
150
+ -1.5,
151
+ -0.6000000238418579,
152
+ -1.5,
153
+ -0.5,
154
+ -0.699999988079071,
155
+ -0.699999988079071,
156
+ -2.216935230032842e-16,
157
+ -2.216935230032842e-16,
158
+ -4.0845591349633594e-18,
159
+ -2.216935230032842e-16,
160
+ -0.4883034825325012,
161
+ 0.1900009959936142,
162
+ -0.5470856428146362,
163
+ -0.34318920969963074,
164
+ -0.35952919721603394,
165
+ -0.35302427411079407,
166
+ -0.4469815790653229,
167
+ -0.6371198296546936,
168
+ -0.7290776371955872,
169
+ -1.0653810501098633,
170
+ -0.8479154706001282,
171
+ -1.0297260284423828,
172
+ -0.42936205863952637,
173
+ -0.5147944092750549,
174
+ -0.16820405423641205,
175
+ -0.045328833162784576,
176
+ -0.13282617926597595,
177
+ 0.7400000095367432,
178
+ -0.5,
179
+ -0.5,
180
+ -1.0,
181
+ -3.138223648071289
182
+ ],
183
+ "action_max": [
184
+ 0.5,
185
+ 0.699999988079071,
186
+ 0.699999988079071,
187
+ 0.0,
188
+ 0.0,
189
+ 0.0,
190
+ 0.0,
191
+ 6.930528109384597e-19,
192
+ 1.108467615016421e-16,
193
+ 1.108467615016421e-16,
194
+ 1.5,
195
+ 1.5,
196
+ 1.0,
197
+ 1.5,
198
+ 0.2557959258556366,
199
+ 0.35884979367256165,
200
+ 0.5090755820274353,
201
+ 0.19132143259048462,
202
+ 0.21249642968177795,
203
+ 0.3992660641670227,
204
+ 0.4283020794391632,
205
+ 0.1457289606332779,
206
+ -0.1900009959936142,
207
+ 0.6150448322296143,
208
+ 0.35468167066574097,
209
+ 0.8703295588493347,
210
+ 0.7531875371932983,
211
+ 0.971237301826477,
212
+ 0.13985762000083923,
213
+ 0.15686897933483124,
214
+ 0.4661160111427307,
215
+ 0.7400000095367432,
216
+ 0.5,
217
+ 0.5,
218
+ 1.0,
219
+ 3.1414895057678223
220
+ ],
221
+ "state_min": [
222
+ -0.02442001923918724,
223
+ -0.0517612099647522,
224
+ -0.0006534014828503132,
225
+ -0.5095356106758118,
226
+ -1.323034405708313,
227
+ -1.3221508264541626,
228
+ -1.3230019807815552,
229
+ -0.5770347714424133,
230
+ -0.4338151812553406,
231
+ -0.6721642017364502,
232
+ -0.0011892789043486118,
233
+ -7.534810038123396e-7,
234
+ -0.001927333534695208,
235
+ -1.075333216249419e-6,
236
+ -0.43650975823402405,
237
+ 0.15721464157104492,
238
+ -0.5489339232444763,
239
+ -0.26434481143951416,
240
+ -0.3508843183517456,
241
+ -0.23784859478473663,
242
+ -0.4281824827194214,
243
+ -0.5803383588790894,
244
+ -0.6814778447151184,
245
+ -1.0344431400299072,
246
+ -0.7932196259498596,
247
+ -1.0205217599868774,
248
+ -0.3445618450641632,
249
+ -0.5986371040344238,
250
+ -0.1348579376935959,
251
+ -0.0017330688424408436,
252
+ -0.1421850621700287,
253
+ 0.7400000095367432,
254
+ 0.0,
255
+ 0.0,
256
+ 0.0,
257
+ 0.0
258
+ ],
259
+ "state_max": [
260
+ 0.47981399297714233,
261
+ 0.6772664189338684,
262
+ 0.6746510863304138,
263
+ 0.0010172375477850437,
264
+ 0.0007091082516126335,
265
+ 0.001881288131698966,
266
+ 0.0011398319620639086,
267
+ 6.141255539660051e-7,
268
+ 0.3043450713157654,
269
+ 6.343479981296696e-7,
270
+ 0.6933000087738037,
271
+ 1.4612544775009155,
272
+ 1.4651201963424683,
273
+ 1.4609057903289795,
274
+ 0.2809508740901947,
275
+ 0.34028318524360657,
276
+ 0.48582184314727783,
277
+ 0.26476219296455383,
278
+ 0.20825636386871338,
279
+ 0.4566418528556824,
280
+ 0.42864030599594116,
281
+ 0.1656116098165512,
282
+ -0.1549365073442459,
283
+ 0.5154499411582947,
284
+ 0.4242899715900421,
285
+ 0.8548054695129395,
286
+ 0.8040095567703247,
287
+ 0.9811649322509766,
288
+ 0.136736661195755,
289
+ 0.195722296833992,
290
+ 0.45781663060188293,
291
+ 0.7400000095367432,
292
+ 0.0,
293
+ 0.0,
294
+ 0.0,
295
+ 0.0
296
+ ],
297
+ "normalize_state": true,
298
+ "pad_action_dim": 36,
299
+ "pad_state_dim": 36
300
+ }
301
+ },
302
+ "root_dir": "/root/workspace/Haoxuan/data/Psi0/data/simple-between-tables-new-training-h264",
303
+ "v5_ardy_sidecar_root": "/root/workspace/Haoxuan/data/Psi0/data/simple-between-tables-new-training-ardy-h264/G1WholebodyLocomotionPickBetweenTablesTeleop-v0",
304
+ "rgb_frame_cache_root": null,
305
+ "train_repo_ids": [
306
+ "G1WholebodyLocomotionPickBetweenTablesTeleop-v0"
307
+ ],
308
+ "val_repo_ids": [
309
+ "G1WholebodyLocomotionPickBetweenTablesTeleop-v0"
310
+ ]
311
+ },
312
+ "model": {
313
+ "resnet_store_path": null,
314
+ "pretrained_action_header_path": "/root/workspace/Haoxuan/data/Psi0/cache/checkpoints/psi0/postpre.1by1.pad36.2601131206.ckpt.he30k",
315
+ "rtc": true,
316
+ "max_delay": 8,
317
+ "action_dim": 36,
318
+ "action_chunk_size": 30,
319
+ "action_exec_horizon": 30,
320
+ "observation_horizon": 1,
321
+ "img_chunk": 1,
322
+ "n_cams": 1,
323
+ "use_obs": "add_token",
324
+ "rsci_variant": "VGQ-AE",
325
+ "rsci_vlm_mode": "frozen",
326
+ "rsci_lora_rank": 8,
327
+ "rsci_lora_alpha": 16.0,
328
+ "psi0_ardy_aux": false,
329
+ "psi0_ardy_aux_weight": 0.1,
330
+ "dropout": 0.1,
331
+ "noise_scheduler": "flow",
332
+ "train_diffusion_steps": 1000,
333
+ "eval_diffusion_steps": 10,
334
+ "share_cam_features": false,
335
+ "early_fusion": false,
336
+ "odim": 36,
337
+ "n_conditions": 0,
338
+ "token_fusion": "concat",
339
+ "loss_w": [
340
+ 0.1,
341
+ 0.2,
342
+ 0.1
343
+ ],
344
+ "time_dim": 256,
345
+ "hidden_dim": 1536,
346
+ "num_blocks": 6,
347
+ "dim_feedforward": 2048,
348
+ "nhead": 24,
349
+ "activation": "gelu",
350
+ "view_feature_dim": 2048,
351
+ "use_film": false,
352
+ "combined_temb": false,
353
+ "use_dit": false,
354
+ "use_depth": false,
355
+ "depth_encoder_name": "defm_vit_s14",
356
+ "depth_encoder_path": null,
357
+ "depth_encoder_dim": 384,
358
+ "depth_encoder_input_size": [
359
+ 196,
360
+ 350
361
+ ],
362
+ "depth_encoder_patch_size": 14,
363
+ "freeze_depth_encoder": true,
364
+ "depth_cross_attention_layers": [
365
+ 0,
366
+ 1,
367
+ 2,
368
+ 3,
369
+ 4,
370
+ 5
371
+ ],
372
+ "depth_fusion_mode": "cross_attention",
373
+ "depth_query_count": 16,
374
+ "depth_query_target_count": 8.0,
375
+ "depth_query_score_dim": 384,
376
+ "depth_query_selection": true,
377
+ "depth_query_use_state": true,
378
+ "depth_query_fuse": false,
379
+ "depth_query_align_depth": false,
380
+ "depth_query_independent_rgb": false,
381
+ "depth_query_concat_streams": false,
382
+ "depth_query_cross_attention_fusion": false,
383
+ "depth_rgb_encoder_path": null,
384
+ "depth_rgb_encoder_type": "dinov2",
385
+ "depth_query_pre_orthogonality": false,
386
+ "depth_query_pre_orthogonality_weight": 0.01,
387
+ "depth_query_post_orthogonality": true,
388
+ "depth_query_resampler_max_tokens": 4096,
389
+ "depth_query_gate_init": 1.0,
390
+ "depth_query_sparsity_init": 0.01,
391
+ "depth_query_sparsity_rate": 0.01,
392
+ "depth_query_sparsity_warmup_steps": 1000,
393
+ "depth_query_recovery_after_steps": 100,
394
+ "depth_query_recovery_weight": 0.01,
395
+ "depth_query_recovery_window_steps": 20,
396
+ "depth_query_recovery_sparsity_scale": 0.0,
397
+ "depth_query_alive_margin": 0.1,
398
+ "depth_query_alive_weight": 1.0,
399
+ "depth_query_selection_weight": 0.01,
400
+ "depth_query_orthogonality_weight": 0.01,
401
+ "v5_enabled": false,
402
+ "v5_film": true,
403
+ "v5_pooling": "attention",
404
+ "v5_ardy": true,
405
+ "v5_hand_state": false,
406
+ "v5_object": true,
407
+ "v5_time_encoding": true,
408
+ "v5_temporal_layers": 0,
409
+ "v5_ardy_weight": 0.1,
410
+ "v5_object_weight": 0.1,
411
+ "v5_intra_modal_dedup": false,
412
+ "v5_intra_modal_dedup_weight": 0.01,
413
+ "depth_gate_init": 0.001,
414
+ "depth_module_lr": 0.0001,
415
+ "action_expert_lr": 0.0001,
416
+ "weight_decay": 0.01,
417
+ "model_name_or_path": "/root/workspace/Haoxuan/data/Psi0/cache/checkpoints/psi0/pre.fast.1by1.2601091803.ckpt.ego200k.he30k",
418
+ "vlm_ckpt_step": null,
419
+ "tune_vlm": false,
420
+ "tune_mm_llm": false,
421
+ "tune_mm_vision": false,
422
+ "tune_mm_mlp": false,
423
+ "gradient_checkpointing": true,
424
+ "lang_backbone_lr": 0.00001,
425
+ "mm_projector_lr": 0.00001,
426
+ "vision_tower_lr": 1e-6,
427
+ "optim": "adamw_torch",
428
+ "model_max_length": 4096,
429
+ "data_flatten": true,
430
+ "data_packing": true,
431
+ "max_pixels": 451584,
432
+ "min_pixels": 12544
433
+ }
434
+ }