Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

This seems to slow things down slightly on Linux. (#10624)

comfyanonymous0f4ef3afa077
diff --git a/comfy/ops.py b/comfy/ops.py
index afe498caa..733bff99d 100644
--- a/comfy/ops.py
+++ b/comfy/ops.py
@@ -35,7 +35,7 @@ def scaled_dot_product_attention(q, k, v, *args, **kwargs):
try:
- if torch.cuda.is_available():
+ if torch.cuda.is_available() and comfy.model_management.WINDOWS:
from torch.nn.attention import SDPBackend, sdpa_kernel
import inspect
if "set_priority" in inspect.signature(sdpa_kernel).parameters: