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.

Only enable cuda malloc on cuda torch. (#9031)

comfyanonymouse78d2304966b
diff --git a/cuda_malloc.py b/cuda_malloc.py
index eb2857c5f..c1d9ae3ca 100644
--- a/cuda_malloc.py
+++ b/cuda_malloc.py
@@ -74,7 +74,8 @@ if not args.cuda_malloc:
module = importlib.util.module_from_spec(spec)
spec.loader.exec_module(module)
version = module.__version__
- if int(version[0]) >= 2: #enable by default for torch version 2.0 and up
+
+ if int(version[0]) >= 2 and "+cu" in version: #enable by default for torch version 2.0 and up only on cuda torch
args.cuda_malloc = cuda_malloc_supported()
except:
pass