Zengtudor/Wan2.1
1
0
Fork 0
mirror of https://github.com/Wan-Video/Wan2.1.git synced 2025-11-04 14:16:57 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Fix flash attention

Browse Source 
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
This commit is contained in:
Emanuele Bugliarello 2025-08-27 11:43:35 +02:00 committed by GitHub
parent 7c81b2f27d
commit ca23a2fc59
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
wan/modules/attention.py
View File

@ -107,7 +107,7 @@ def flash_attention(
max_seqlen_k=lk, max_seqlen_k=lk,
softmax_scale=softmax_scale, softmax_scale=softmax_scale,
causal=causal, causal=causal,
deterministic=deterministic)[0].unflatten(0, (b, lq)) deterministic=deterministic).unflatten(0, (b, lq))
else: else:
assert FLASH_ATTN_2_AVAILABLE assert FLASH_ATTN_2_AVAILABLE
x = flash_attn.flash_attn_varlen_func( x = flash_attn.flash_attn_varlen_func(