提交图

56 次代码提交

作者 SHA1 备注 提交日期
ee4433d812 actuator network 2023-11-18 20:50:56 -08:00
71c46d8649 fix compress_data.py 2023-11-14 17:13:33 -08:00
b7e387a9b9 Merge branch 'main' of github.com:MarkFzp/act-plus-plus into main 2023-11-14 14:43:05 -08:00
8f66f3bc88 use mobile-aloha2 2023-11-14 14:38:11 -08:00
79e5032002 Diffusion policy 2023-11-14 11:02:33 -08:00
61decb1e13 add compress to postprocess_episodes.py 2023-11-11 19:06:55 -08:00
5376392e66 dataloader __getitem__ supported compressed 2023-11-11 18:16:15 -05:00
329a237290 fix video viz bgr 2023-11-10 21:12:11 -08:00
d89430f987 compress_data.py 2023-11-05 00:54:36 -07:00
4cec62314b default deploy with c=0.19 calibration, add postprocess_base_action to utils.py for shrinked angular vel 2023-11-02 23:46:29 -07:00
5ab307d5c4 Merge branch 'main' of github.com:MarkFzp/act-plus-plus into main 2023-11-01 18:54:11 -07:00
b401b8f9d6 align.py for cam and arm symmetry calibration 2023-11-01 18:54:06 -07:00
902dae2236 remove wandb when is_eval, change smooth action to conv 5 2023-11-01 21:53:04 -04:00
60d0da09f8 Merge branch 'mirror' into main 2023-11-01 18:08:43 -07:00
bc860dd58c fix smooth_base_action in utils.py 2023-11-01 15:32:34 -07:00
e8aa7976ba update postprocess_episodes for base_action, skip_mirrored_data option 2023-11-01 03:54:06 -04:00
7678af7021 no encoder option 2023-10-31 13:23:35 -07:00
dc756964a2 Merge branch 'mirror' of github.com:MarkFzp/act-plus-plus into mirror 2023-10-31 02:41:42 -04:00
02a6a2e4e5 add resume_ckpt_path option, sim_insertion_scripted_mirror task 2023-10-31 02:41:36 -04:00
10c445adb0 small 2023-10-30 22:51:11 -07:00
6afe28b795 load pretrain flag 2023-10-30 02:35:34 -07:00
cbc01948c8 fix pretraining on all data 2023-10-29 23:30:52 -07:00
128701a22f fix bug and add mirroring 2023-10-29 23:02:40 -07:00
588727d415 fix env shaking at first step 2023-10-29 12:34:26 -07:00
33e4c86341 improved 2023-10-28 23:44:40 -07:00
8119013853 debug smooth_base_action astype(np.float32), train_ratio = 0.99 2023-10-28 16:10:19 -07:00
d6bb889d7d Merge branch 'main' of github.com:MarkFzp/act-plus-plus into main 2023-10-26 00:57:42 -07:00
d0aa344ff0 remove preprocess base_action calibration, add postprocess option in eval_bc() 2023-10-26 00:56:33 -07:00
8f22418a7b Merge branch 'main' of github.com:MarkFzp/act-plus-plus into main 2023-10-26 01:44:36 -04:00
654248d96f fix in-place calibrate_linear_vel 2023-10-26 01:44:31 -04:00
3474cc38e7 Merge branch 'main' of github.com:MarkFzp/act-plus-plus into main 2023-10-25 21:56:24 -07:00
ca82042c1c make imitate_episodes.py runnable for mobile_aloha eval 2023-10-25 21:56:15 -07:00
b75a32467b add linear vel calibration, refactor base_vel proprocessing in utils.py 2023-10-25 15:10:08 -04:00
26a457ab4e add conv filter to EpisodicDataset in utils.py, save ckpt every 1000 steps 2023-10-24 21:56:33 -07:00
7f62bface3 fix action dim for real world 2023-10-23 17:22:17 -07:00
89cea60069 fix vq 2023-10-23 19:33:15 -04:00
b37712fd2f vq init 2023-10-13 17:57:59 -04:00
694c606940 remove clip ceiling 2023-09-04 23:12:06 -07:00
55098451f8 Update README.md for python version 2023-08-30 18:30:28 -07:00
dfe6c7f5ff typo in VAE encoder's joint projection 2023-06-23 13:30:39 -07:00
f74f0db292 add MIT license 2023-06-07 13:16:18 -07:00
75d93f1e38 small typo 2023-05-31 20:42:43 -07:00
f2d94d016e Update README.md 2023-05-16 22:45:41 -07:00
5b3acceada add tuning tips to README.md 2023-05-04 23:36:16 -07:00
dee143171e Mention temporal ensemble flag in README.md 2023-05-04 15:41:27 -07:00
57d9206b38 Change network architecture to have joints as ACT encoder input
- does not affect performance for position control
- not backward compatible: policies trained before this commit will not load because of additional params.
2023-04-14 11:50:27 -07:00
b16ccad5ff Merge pull request #2 from cheng-chi/cchi/reproduce
Small fixes for dependencies
2023-04-11 15:17:35 -07:00
589f691175 updated README 2023-04-11 17:47:25 -04:00
ac37128945 added gitignore conda and fixed h5py_cache dependency 2023-04-11 15:19:07 -04:00
0acd1fa4f2 added conda env 2023-04-11 15:00:56 -04:00