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.

Disable bf16 on AMD GPUs that don't support it.

comfyanonymouse2919d38b4a7
diff --git a/comfy/model_management.py b/comfy/model_management.py
index fb924f432..9f6522967 100644
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@@ -1106,6 +1106,11 @@ def should_use_bf16(device=None, model_params=0, prioritize_performance=True, ma
if is_ascend_npu():
return True
+ if is_amd():
+ arch = torch.cuda.get_device_properties(device).gcnArchName
+ if arch in ["gfx1030", "gfx1031", "gfx1010", "gfx1011", "gfx1012", "gfx906", "gfx900", "gfx803"]: # RDNA2 and older don't support bf16
+ return False
+
props = torch.cuda.get_device_properties(device)
if props.major >= 8:
return True