XPU Not working "No backend can handle 'dequantize_per_tensor_fp8': eager: x: device xpu not in {'cuda', 'cpu'}"

#3
by AI-Joe-git - opened

No backend can handle 'dequantize_per_tensor_fp8': eager: x: device xpu not in {'cuda', 'cpu'}

System Information

  • ComfyUI Version: 0.7.0
  • Arguments: main.py --output-directory C:\Users\uscha\Documents\AI-Playground\media --front-end-version Comfy-Org/ComfyUI_frontend@latest --max-upload-size 500
  • OS: win32
  • Python Version: 3.11.0 (main, Oct 24 2022, 18:26:48) [MSC v.1933 64 bit (AMD64)]
  • Embedded Python: false
  • PyTorch Version: 2.11.0.dev20251215+xpu

Devices

  • Name: xpu:0 Intel(R) Arc(TM) 140V GPU
    • Type: xpu
    • VRAM Total: 27230117888
    • VRAM Free: 27230116864
    • Torch VRAM Total: 2097152
    • Torch VRAM Free: 2096128

Do you have Kitchen? I don't think it supports XPU, does it?

Yes, I do, but it doesn’t even switch to CPU if CUDA is not available. My only hope is to wait for a proper GGUF version that might resolve the issue. I've successfully run inference with the --cpu flag, but as you can imagine, it's so slow that it’s not practical.

UPDATE: It works now, and it's incredibly fast !

Sign up or log in to comment