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 run text encoders on xpu because there are issues.

comfyanonymous0966d3ce823d
diff --git a/comfy/model_management.py b/comfy/model_management.py
index b663e8f59..e38ef4eea 100644
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@@ -451,6 +451,8 @@ def text_encoder_device():
if args.gpu_only:
return get_torch_device()
elif vram_state == VRAMState.HIGH_VRAM or vram_state == VRAMState.NORMAL_VRAM:
+ if is_intel_xpu():
+ return torch.device("cpu")
if should_use_fp16(prioritize_performance=False):
return get_torch_device()
else: