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.

Don't use is_bf16_supported to check for fp16 support.

comfyanonymous66e28ef45c02
diff --git a/comfy/model_management.py b/comfy/model_management.py
index aa40c502a..a8dc91b9e 100644
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@@ -722,10 +722,13 @@ def should_use_fp16(device=None, model_params=0, prioritize_performance=True, ma
if is_intel_xpu():
return True
- if torch.cuda.is_bf16_supported():
+ if torch.version.hip:
return True
props = torch.cuda.get_device_properties("cuda")
+ if props.major >= 8:
+ return True
+
if props.major < 6:
return False