Skip to content

Fix Flash Attention 3 interface for new FA3 return format #13173

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
veeceey wants to merge 2 commits into
base: main
Choose a base branch
from
+10 −2
Open

Fix Flash Attention 3 interface for new FA3 return format #13173

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d7e2c07
Fix Flash Attention 3 interface compatibility for new FA3 versions
veeceey Feb 23, 2026
c0961d5
Simplify _wrapped_flash_attn_3 return unpacking
veeceey Feb 24, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 10 additions & 2 deletions src/diffusers/models/attention_dispatch.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -676,7 +676,7 @@ def _wrapped_flash_attn_3(
) -> tuple[torch.Tensor, torch.Tensor]:
# Hardcoded for now because pytorch does not support tuple/int type hints
window_size = (-1, -1)
out, lse, *_ = flash_attn_3_func(
result = flash_attn_3_func(
q=q,
k=k,
v=v,
Expand All @@ -693,7 +693,9 @@ def _wrapped_flash_attn_3(
pack_gqa=pack_gqa,
deterministic=deterministic,
sm_margin=sm_margin,
return_attn_probs=True,
)
out, lse, *_ = result
lse = lse.permute(0, 2, 1)
return out, lse

Expand Down Expand Up @@ -2623,7 +2625,7 @@ def _flash_varlen_attention_3(
key_packed = torch.cat(key_valid, dim=0)
value_packed = torch.cat(value_valid, dim=0)

out, lse, *_ = flash_attn_3_varlen_func(
result = flash_attn_3_varlen_func(
q=query_packed,
k=key_packed,
v=value_packed,
Expand All @@ -2633,7 +2635,13 @@ def _flash_varlen_attention_3(
max_seqlen_k=max_seqlen_k,
softmax_scale=scale,
causal=is_causal,
return_attn_probs=return_lse,
)
if isinstance(result, tuple):
out, lse, *_ = result
else:
out = result
lse = None
out = out.unflatten(0, (batch_size, -1))

return (out, lse) if return_lse else out
Expand Down
Loading