Commit graph

460 commits

Author SHA1 Message Date
Shen-Chenhui
bcb997502e enable disc 2024-05-02 17:28:45 +08:00
Shen-Chenhui
b35b933841 add disc 2024-05-02 16:32:19 +08:00
Zangwei Zheng
00e21e9fed [wip] adapt multi-res 2024-05-02 15:27:03 +08:00
zhengzangw
02904b4f29 updated 2024-05-02 03:46:01 +00:00
zhengzangw
f7a4fcfc3f updated 2024-04-30 09:46:27 +00:00
zhengzangw
b89c072eb5 [fix] micro batch size not for training 2024-04-30 09:20:08 +00:00
zhengzangw
ace37cf7c7 refactor 2024-04-30 08:13:20 +00:00
zhengzangw
41e276f5ef updated 2024-04-30 06:19:40 +00:00
周钰坤
bf633f3e68 Update pixart.py 2024-04-30 13:53:17 +08:00
周钰坤
34711b7bd2 update 2024-04-30 13:37:25 +08:00
周钰坤
59591d8afc update 2024-04-30 13:36:41 +08:00
zhengzangw
b2f302b379 [wip] try vae 2024-04-29 17:02:22 +00:00
Shen-Chenhui
e3b84247c1 disable t-conv for t=1 2024-04-29 20:06:18 +08:00
Shen-Chenhui
c59b499085 Merge branch 'vae-clean-new' of https://github.com/hpcaitech/Open-Sora-dev into vae-clean-new 2024-04-29 18:10:03 +08:00
Shen-Chenhui
c631358bbf add z nll loss 2024-04-29 18:09:42 +08:00
zhengzangw
9632b5bcf8 merged 2024-04-29 10:07:09 +00:00
zhengzangw
5c8139521a update 2024-04-29 10:02:42 +00:00
Shen-Chenhui
b78b469420 fix merge 2024-04-29 17:58:59 +08:00
Shen-Chenhui
f81d8648cf move out losses to different file 2024-04-29 17:53:03 +08:00
zhengzangw
a2dc903863 update 2024-04-29 09:41:39 +00:00
Shen-Chenhui
4d69671663 clean up code 2024-04-29 17:18:45 +08:00
zhengzangw
1171e5b6f9 update config 2024-04-29 07:27:15 +00:00
Shen-Chenhui
98de78910d config 2024-04-29 15:21:54 +08:00
Shen-Chenhui
7811b8e99a add Nlayer 3D discriminator 2024-04-29 14:18:57 +08:00
Shen-Chenhui
069ffcc4e3 use 3 layer disc 2024-04-29 14:03:18 +08:00
zhengzangw
f3dcb4d6fb complete 3b backbone 2024-04-29 06:00:14 +00:00
Shen-Chenhui
8b465b04c5 add vae subfolder support 2024-04-29 11:32:39 +08:00
Tianyi
07cbe11cb6 Merge branch 'main' of https://github.com/hpcaitech/Open-Sora-dev into dev/Rflow 2024-04-28 05:59:20 +00:00
zhengzangw
84465c0e44 add config for PixArt 1.6B 2024-04-27 13:41:21 +00:00
zhengzangw
7b08fb289c [fix] pixart-sigma 2024-04-27 13:19:00 +00:00
zhengzangw
d6a16a858e [wip] multiple resolution 2024-04-27 11:43:15 +00:00
zhengzangw
f551321e49 Merge branch 'main' into dev/v1.2 2024-04-27 11:40:56 +00:00
Shen-Chenhui
6fb4e3cd22 debug 2024-04-27 17:18:19 +08:00
Shen-Chenhui
c9785342c2 add evals 2024-04-27 16:14:11 +08:00
Zheng Zangwei (Alex Zheng)
8f109b7117 fix qknorm bug (#80) 2024-04-27 15:52:17 +08:00
YuKun Zhou
fddd08243b Update ckpt_utils.py (#79)
* Update ckpt_utils.py

* update

* Update 1x256x256.py

* update

* update

* Update blocks.py

* update

* Update 1x2048x2048.py
2024-04-27 15:43:50 +08:00
Shen-Chenhui
38c46ac721 fix bug 2024-04-27 15:25:03 +08:00
Shen Chenhui
b5a3334caf fix dataset loading 2024-04-27 14:52:52 +08:00
Shen Chenhui
3e7ba9a46d use modulelist 2024-04-27 14:02:20 +08:00
zhengzangw
478b585024 [wip] debug vae 2024-04-26 07:27:26 +00:00
zhengzangw
a4652a8aef Merge branch 'main' into vae 2024-04-26 06:56:22 +00:00
zhengzangw
7fcf0c6bfd fix qknorm bug 2024-04-26 03:59:50 +00:00
Shen-Chenhui
576b44d98e add lecam storage support 2024-04-26 10:34:00 +08:00
Tianyi
006ab8f35a Merge branch 'main' of https://github.com/hpcaitech/Open-Sora-dev into dev/Rflow 2024-04-25 03:16:54 +00:00
Frank Lee
367ae02e7c updated gradio app (#74) 2024-04-24 22:30:36 +08:00
zhengzangw
44e618bbbc a bunch of update 2024-04-24 09:23:24 +00:00
Shen-Chenhui
9683ec66c7 inference script update 2024-04-24 17:04:53 +08:00
Shen-Chenhui
0138be6e21 grad penalty expr 2024-04-24 14:52:47 +08:00
zhengzangw
9e0b314dab update config 2024-04-24 02:33:27 +00:00
Shen-Chenhui
988bc3bb65 config lc gp 2024-04-24 10:03:58 +08:00
Zangwei Zheng
e009e89e1a add start from scratch 2024-04-23 19:24:22 +08:00
Shen-Chenhui
48e3a84ba6 enable r1 gradient penalty 2024-04-23 17:51:50 +08:00
zhengzangw
cf95f41002 update docs 2024-04-23 09:26:10 +00:00
Shen-Chenhui
d565477680 disable non-master print 2024-04-23 14:43:02 +08:00
Shen-Chenhui
d0d7da36ea silence lpips 2024-04-23 11:44:33 +08:00
Shen-Chenhui
2ac952bb71 8 GPU training config 2024-04-23 11:30:40 +08:00
Shen-Chenhui
486593f137 LeCam Loss working 2024-04-23 10:58:09 +08:00
Shen-Chenhui
16b0149fc4 lecam config 2024-04-23 09:59:12 +08:00
Shen-Chenhui
f4fe9b5eca add lecam 2024-04-23 09:22:05 +08:00
Zangwei Zheng
3944e93065 update training config 2024-04-22 12:05:59 +08:00
zhengzangw
8931efccd7 update config 2024-04-22 03:43:31 +00:00
Zangwei Zheng
fea6aba655 hotfix config args 2024-04-21 19:50:51 +08:00
zhengzangw
da345719c7 use qk-norm 2024-04-21 11:40:20 +00:00
zhengzangw
2b42ca82d8 [wip] prompt with json 2024-04-21 11:07:59 +00:00
zhengzangw
6073af6444 merge video edit 2024-04-20 15:47:48 +00:00
Zheng Zangwei (Alex Zheng)
341b12f9bf Dev/v1.0.1 (#58)
* update (#57)

* update

* update datautil

* add VBench prompt

* update eval

* update eval

* update intepolation

* add vbench eval

* Dev/sdedit implementation (#56)

* Update utils.py

* update

* update

* update

---------

Co-authored-by: YuKun Zhou <90625606+1zeryu@users.noreply.github.com>
2024-04-20 21:23:10 +08:00
zhengzangw
d262f14a24 update 2024-04-20 02:43:38 +00:00
Zangwei Zheng
74aade586a [exp] update 2024-04-19 17:56:02 +08:00
Shen-Chenhui
96a42e08db config 2024-04-19 17:49:50 +08:00
Zangwei Zheng
15f9d702ed [exp] update 2024-04-19 17:43:44 +08:00
Zangwei Zheng
1877a14abc load ema strict=False 2024-04-19 16:55:03 +08:00
Zangwei Zheng
8bf34f4418 add local_files_only 2024-04-19 16:47:26 +08:00
Shen-Chenhui
4c46972b64 debug 2024-04-19 15:59:21 +08:00
Shen-Chenhui
19187b70f5 allow specifying outputs in command line 2024-04-19 15:56:04 +08:00
Zangwei Zheng
dce2ef4a1c [exp] new config 2024-04-19 15:21:00 +08:00
Shen-Chenhui
78caeea999 add support and config for z channel 4 2024-04-19 15:14:53 +08:00
Shen-Chenhui
b16001e5e3 debug 2024-04-19 14:28:53 +08:00
Shen-Chenhui
89a58b8e15 debug 2024-04-19 14:23:57 +08:00
Shen-Chenhui
1c3e910fc6 allow loading of discriminator in inference 2024-04-19 14:19:29 +08:00
Shen-Chenhui
6581556997 pipeline config 2024-04-19 13:36:47 +08:00
Zheng Zangwei (Alex Zheng)
d2a782efac Dev/fps (#55)
* support fps

* update fps
2024-04-19 13:18:52 +08:00
Zangwei Zheng
0e6c15d50b Merge branch 'main' into dev/v1.0.1 2024-04-19 12:48:22 +08:00
Zheng Zangwei (Alex Zheng)
069ea0d687 Feat/fast bucket (#54)
* [wip] bucket

* [bug] not parallel

* update eval

* update sample.sh

* accelerate bucket build with pandarallel
2024-04-19 11:42:02 +08:00
tianyi
0045d8b7b0 fix 2024-04-19 11:18:29 +08:00
tianyi
0d98c1d02d Merge branch 'dev/v1.0.1' of github.com:hpcaitech/Open-Sora-dev into dev/Rflow 2024-04-19 11:18:01 +08:00
Shen-Chenhui
41a600b61b inference config 2024-04-19 10:47:09 +08:00
Shen-Chenhui
0ad337178a debug 2024-04-19 10:29:25 +08:00
Shen-Chenhui
34be856fdc add disc initialization code 2024-04-19 09:39:23 +08:00
corey-lambda
c0f3c3a9ca Create accleration/shardformer/__init__.py (#250) 2024-04-18 23:34:47 +08:00
7sunarni
d90c049a2f feat(ckpt_utils): support HF_MIRROR (#264)
Signed-off-by: 7sunarni <710720732@qq.com>
2024-04-18 23:31:40 +08:00
zhengzangw
2539a610d5 update mask strategy 2024-04-18 08:13:53 +00:00
Zheng Zangwei (Alex Zheng)
1f99b9ba4b complete eval pipeline (#53) 2024-04-18 15:49:14 +08:00
Shen-Chenhui
1b86c24de6 remove breakpoint 2024-04-18 15:22:22 +08:00
Shen-Chenhui
60892cd090 debug 2024-04-18 15:11:25 +08:00
Shen-Chenhui
246eddecdd enable pipeline vae training 2024-04-18 14:13:56 +08:00
Shen-Chenhui
f867ebc819 debug 2024-04-18 13:29:59 +08:00
Shen-Chenhui
8bf4a0fa77 add disable space in vae v2 2024-04-18 12:23:02 +08:00
Shen-Chenhui
e241ecbbe1 use default discriminator 2024-04-18 09:32:17 +08:00
Shen-Chenhui
5084d4e4f9 disc 2024-04-17 18:12:56 +08:00
Zangwei Zheng
ee2e4083e4 [feat] update eval 2024-04-17 17:35:44 +08:00
Zangwei Zheng
380ea6cce8 Merge branch 'dev/v1.0.1' of https://github.com/hpcaitech/Open-Sora-dev into dev/v1.0.1 2024-04-17 17:05:12 +08:00
Zangwei Zheng
8cb6d2f0bd [feat] update eval 2024-04-17 17:05:08 +08:00
FrankLeeeee
6ee754bef9 print error data path 2024-04-17 16:45:18 +08:00
Shen-Chenhui
de0199c6b2 debug 2024-04-16 17:30:03 +08:00
Shen-Chenhui
235d195dac add logvar 2024-04-16 17:25:48 +08:00
Zheng Zangwei (Alex Zheng)
a72e59610f update extending ynull (#51) 2024-04-16 16:45:06 +08:00
tianyi
2d1590b582 fix bugs, runable 2024-04-16 16:07:30 +08:00
Shen-Chenhui
afd3f823d4 debug 2024-04-16 15:00:31 +08:00
tianyi
62d96a2ced Merge branch 'dev/v1.0.1' of github.com:hpcaitech/Open-Sora-dev into dev/Rflow 2024-04-15 19:42:45 +08:00
Shen-Chenhui
af1b1e484d reimplement blurpool 2024-04-15 17:47:33 +08:00
Zangwei Zheng
50ad9a1ca5 small update 2024-04-15 17:36:20 +08:00
Shen-Chenhui
686829969f debug 2024-04-15 15:13:29 +08:00
Zangwei Zheng
63e86f6dd6 hotfix inference 2024-04-15 13:44:30 +08:00
Shen-Chenhui
c71e04daaa add lecam and gradient penalty loss to discriminator 2024-04-15 11:13:39 +08:00
tianyi
d7a54fa736 Merge branch 'dev/v1.0.1' of github.com:hpcaitech/Open-Sora-dev into dev/Rflow 2024-04-15 10:45:59 +08:00
Zangwei Zheng
3b85effe5a a bunch of update for data 2024-04-13 15:44:24 +08:00
Shen-Chenhui
79bff13099 debug 2024-04-13 10:51:56 +08:00
Shen-Chenhui
fb0f59171c debug 2024-04-13 09:36:39 +08:00
Shen-Chenhui
8aad3d6de2 debug 2024-04-12 18:35:28 +08:00
Shen-Chenhui
c3763cb556 debug 2024-04-12 18:32:24 +08:00
Shen-Chenhui
5c046c4771 debug 2024-04-12 18:22:35 +08:00
Shen-Chenhui
735e9e722b debug 2024-04-12 18:18:06 +08:00
Shen-Chenhui
d53245a98c debug 2024-04-12 18:14:18 +08:00
Shen-Chenhui
17d6b58178 debug 2024-04-12 18:08:42 +08:00
Shen-Chenhui
eae76b5a04 debug 2024-04-12 18:00:52 +08:00
Shen-Chenhui
de9fffacfd gan training 2024-04-12 17:57:35 +08:00
Shen-Chenhui
dfb6af84d4 debug 2024-04-12 15:21:40 +08:00
Shen-Chenhui
a97c70fab2 debug 2024-04-12 15:20:50 +08:00
Shen-Chenhui
e004d43f69 gan optimizer 2024-04-12 15:10:19 +08:00
Shen-Chenhui
ab59033f20 gan 2024-04-12 12:42:02 +08:00
Shen-Chenhui
19fb131e49 debug 2024-04-11 17:54:11 +08:00
Shen-Chenhui
f0756dd7a6 debug 2024-04-11 17:49:15 +08:00
Shen-Chenhui
30353e1351 debug 2024-04-11 17:46:47 +08:00
Shen-Chenhui
a5015c46c8 debug 2024-04-11 17:19:05 +08:00
Shen-Chenhui
5abb9239fb debug 2024-04-11 17:04:52 +08:00
Zangwei Zheng
4d338419a7 add cache dir 2024-04-11 16:46:57 +08:00
Shen-Chenhui
89a53d423c debug 2024-04-11 16:44:02 +08:00
Shen-Chenhui
dbe1c4cf2f debug 2024-04-11 16:33:22 +08:00
Shen-Chenhui
4741d952d8 debug 2024-04-11 16:25:06 +08:00
Shen-Chenhui
158bbfa2f0 debug 2024-04-11 16:00:42 +08:00
Shen-Chenhui
38c1572e66 debug 2024-04-11 15:51:23 +08:00
Shen-Chenhui
c0632b3fb3 debug 2024-04-11 15:50:11 +08:00
Shen-Chenhui
772a52aafa debug 2024-04-11 15:46:27 +08:00
Shen-Chenhui
a769ca6d86 debug 2024-04-11 15:42:37 +08:00
Shen-Chenhui
1dd6170c89 debug 2024-04-11 15:39:51 +08:00
Shen-Chenhui
f8a1cc34b2 debug 2024-04-11 15:38:40 +08:00
Shen-Chenhui
f6c79a75c2 debug 2024-04-11 15:37:22 +08:00
Shen-Chenhui
4cb946c547 debug 2024-04-11 15:35:55 +08:00
Shen-Chenhui
ccb92cd1ab debug 2024-04-11 15:34:11 +08:00
Zangwei Zheng
fff002bede [feat] support parquet 2024-04-11 15:28:49 +08:00
Shen-Chenhui
9cb8e42622 add kl loss 2024-04-11 15:25:50 +08:00
Shen-Chenhui
63f3737eb0 add simple discriminator 2024-04-11 15:14:16 +08:00
Zangwei Zheng
da1038ca5c a bunch of update 2024-04-11 11:48:06 +08:00
ChangeFWorld
6fa92af355 rflow coarse finish, no check yet 2024-04-11 11:31:21 +08:00
Shen-Chenhui
dc63a2b4e5 debug 2024-04-11 11:28:05 +08:00
Shen-Chenhui
ca1d5863c3 add perceptual loss 2024-04-11 11:23:40 +08:00
Shen-Chenhui
e5f881ca1a debug 2024-04-11 10:53:14 +08:00
Shen-Chenhui
ee58ec16b4 add image encoding support 2024-04-11 10:50:23 +08:00
ChangeFWorld
9a14b9e23c rflow scheduler unfinished(40%) 2024-04-10 18:27:43 +08:00
ChangeFWorld
d00f6810bb typo fix 2024-04-10 10:07:21 +08:00
Shen-Chenhui
fef0bf875b debug 2024-04-09 18:28:19 +08:00
Shen-Chenhui
84c6918e2e debug 2024-04-09 18:26:21 +08:00
Shen-Chenhui
5064626844 debug padding 2024-04-09 18:24:39 +08:00
Shen-Chenhui
1bbef18f15 strides debug 2024-04-09 18:14:25 +08:00
Shen-Chenhui
d8666a73c4 register module 2024-04-09 18:04:08 +08:00
Shen-Chenhui
0fdc6e9935 config 2024-04-09 17:56:42 +08:00
Shen-Chenhui
a4025b0ea7 magvit v2 enc dec arc 2024-04-09 17:49:01 +08:00
Zangwei Zheng
e00f93d111 [exp] update 2024-04-09 13:55:55 +08:00
Zangwei Zheng
71a48c5d27 training config 2024-04-09 10:23:28 +08:00
Shen-Chenhui
ae6bffb8c5 restore train vae script 2024-04-09 09:34:59 +08:00
Zangwei Zheng
e1d98678be [fix] dataset ext judge 2024-04-08 20:44:01 +08:00
Zangwei Zheng
8d3f0f8282 [fix] dataset ext judge 2024-04-08 20:42:38 +08:00
Shen-Chenhui
47b51ba07d debug 2024-04-08 16:53:18 +08:00
Shen-Chenhui
c6181ebdfd debug 2024-04-08 16:36:54 +08:00
Shen-Chenhui
fa0ca3983e debug inference code 2024-04-08 15:40:14 +08:00
Shen-Chenhui
4b27448a49 debug 2024-04-08 14:22:21 +08:00
Shen-Chenhui
ad02d28d9c debug inference running loss 2024-04-08 10:22:07 +08:00
Zangwei Zheng
50c0a35658 [fix] bucket 2024-04-08 02:28:14 +08:00
Zangwei Zheng
2685f2930e fix bucket index 2024-04-07 19:48:44 +08:00
Zangwei Zheng
02b6335628 fix bucket index 2024-04-07 19:35:30 +08:00
Zangwei Zheng
882225897b fix bucket index 2024-04-07 16:28:44 +08:00
Hongxin Liu
1182003c9b [feature] accelerate iddpm (#39) 2024-04-07 14:30:38 +08:00
Shen-Chenhui
88be217a13 debug 2024-04-05 16:25:32 +08:00
Shen-Chenhui
e3584b4e43 debug 2024-04-05 16:20:39 +08:00
Shen-Chenhui
9091419b28 save vae to model dir 2024-04-05 16:07:23 +08:00
Shen-Chenhui
44b8142b4f add wrapper for pretrained loading 2024-04-05 16:01:24 +08:00
Shen-Chenhui
c7506e2461 adding pretrained vae 2024-04-04 15:54:55 +08:00
Shen-Chenhui
a656421c83 debug from_pretrained in progress 2024-04-04 15:41:15 +08:00
Shen-Chenhui
d08d52ea03 config add is_vae 2024-04-04 15:23:57 +08:00
Shen-Chenhui
60eaffed4a debug 2024-04-04 15:11:42 +08:00
Shen-Chenhui
4250cfa9a2 debug 2024-04-04 15:10:11 +08:00
Shen-Chenhui
e9776584e4 allow vae not to have prompt file 2024-04-04 15:06:36 +08:00
Shen-Chenhui
9dc9afb254 add data path to inference config 2024-04-04 14:39:04 +08:00
Shen-Chenhui
dfdf8db491 add inference config
add readme
complete inference script
TODO: DEBUG inference
2024-04-04 14:11:25 +08:00
Zangwei Zheng
6762476261 update mask & config 2024-04-03 17:25:41 +08:00
Zangwei Zheng
12dcbeee8e update csvutil & mask 2024-04-03 15:15:04 +08:00
Shen-Chenhui
b6c39873ee finish discriminator arch
use einops.rearrange in vae model_utils
add get_latent_size in vae_3d
inference vae_3d WIP
2024-04-03 10:29:01 +08:00
Hongxin Liu
c618f77bdc [feature] move t5 text processing to csvutil (#22)
* [feature] move t5 text processing to csvutil

* polish code
2024-04-02 14:56:44 +08:00
Zheng Zangwei (Alex Zheng)
7612d22fc6 Exp/image mix (#20)
* [exp] image mixed training

* [exp] add batch info

* [exp] launch

* update num_step_per_epoch

* [feat] verify image mix training
2024-04-02 13:35:09 +08:00
Shen-Chenhui
aa733aa1d7 debug 2024-04-02 11:21:01 +08:00