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.

Enable async offload by default for AMD. (#11534)

comfyanonymous2943093a5310
diff --git a/comfy/model_management.py b/comfy/model_management.py
index 1889ab0ac..e5554e225 100644
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@@ -1019,8 +1019,8 @@ NUM_STREAMS = 0
if args.async_offload is not None:
NUM_STREAMS = args.async_offload
else:
- # Enable by default on Nvidia
- if is_nvidia():
+ # Enable by default on Nvidia and AMD
+ if is_nvidia() or is_amd():
NUM_STREAMS = 2
if args.disable_async_offload: