Skip to content

[pull] master from Comfy-Org:master #1003

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

Merged
pull merged 1 commit into from
Jul 30, 2026
Merged

[pull] master from Comfy-Org:master #1003

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
7 changes: 5 additions & 2 deletions comfy/ops.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,7 +41,7 @@ def scaled_dot_product_attention(q, k, v, *args, **kwargs):


try:
if torch.cuda.is_available() and comfy.model_management.WINDOWS:
if torch.cuda.is_available():
from torch.nn.attention import SDPBackend, sdpa_kernel
import inspect
if "set_priority" in inspect.signature(sdpa_kernel).parameters:
Expand All @@ -51,7 +51,10 @@ def scaled_dot_product_attention(q, k, v, *args, **kwargs):
SDPBackend.MATH,
]

SDPA_BACKEND_PRIORITY.insert(0, SDPBackend.CUDNN_ATTENTION)
if comfy.model_management.WINDOWS:
SDPA_BACKEND_PRIORITY.insert(0, SDPBackend.CUDNN_ATTENTION)
else:
SDPA_BACKEND_PRIORITY.insert(1, SDPBackend.CUDNN_ATTENTION)

def scaled_dot_product_attention(q, k, v, *args, **kwargs):
if q.nelement() < 1024 * 128: # arbitrary number, for small inputs cudnn attention seems slower
Expand Down
Loading