Zengtudor/Wan2.1
1
0
Fork 0
mirror of https://github.com/Wan-Video/Wan2.1.git synced 2025-11-05 06:29:14 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
46 Commits 6 Branches 0 Tags 22 MiB
ca23a2fc59
Commit Graph

11 Commits

Author SHA1 Message Date
Emanuele Bugliarello
ca23a2fc59
Fix flash attention 
fa3 latest version changed the return shape of the varlen func to be consistent w fa2. this pr fixes the fa3 attention call as done in https://github.com/Wan-Video/Wan2.2/pull/64
 2025-08-27 11:43:35 +02:00
Ang Wang
76e9427657
Format the code (#402) 
* isort the code

* format the code

* Add yapf config file

* Remove torch cuda memory profiler
 2025-05-16 12:35:38 +08:00
Zhen Han
c709fcf0e7
fix vace size (#397) 2025-05-14 22:01:45 +08:00
Ang Wang
18d53feb7a
[feature] Add VACE (#389) 
* Add VACE

* Support training with multiple gpus

* Update default args for vace task

* vace block update

* Add vace exmaple jpg

* Fix dist vace fwd hook error

* Update vace exmample

* Update vace args

* Update pipeline name for vace

* vace gradio and Readme

* Update vace snake png

---------

Co-authored-by: hanzhn <han.feng.jason@gmail.com>
 2025-05-14 20:44:25 +08:00
yupeng1111
df44622e72
[feature] Wan2.1-FLF2V-14B (#338) 
Co-authored-by: 澎鹏 <shiyupeng.syp@taobao.com>
 2025-04-17 21:56:46 +08:00
jiangzeyinzi
d6434cf8ef
To support system prompt as a input parameter (#280) 2025-03-28 11:51:31 +08:00
Wenting Shen
bc3249d61c
add free stroage api for FSDP (#277) 
Signed-off-by: wenting.swt <wenting.swt@alibaba-inc.com>
 2025-03-26 15:49:37 +08:00
Yingda Chen
d0a086800a
Update prompt_extend.py 2025-03-06 13:11:22 +08:00
Alan May
82c18d91a1
fix: correct Chinese ISO language code typo (#10) 
Fix incorrect Chinese language ISO code abbreviation
from CH to ZH
 2025-03-03 17:04:13 +08:00
Adrian Corduneanu
0e3c42a830
Update text2video.py to reduce GPU memory by emptying cache (#44) 
* Update text2video.py to reduce GPU memory by emptying cache

If offload_model is set, empty_cache() must be called after the model is moved to CPU to actually free the GPU. I verified on a RTX 4090 that without calling empty_cache the model remains in memory and the subsequent vae decoding never finishes.

* Update text2video.py only one empty_cache needed before vae decode
 2025-02-26 18:56:57 +08:00
WanX-Video-1
65386b2e03 init upload 2025-02-25 22:07:47 +08:00