Skip to content

v1.0.1 — Intel Arc Build Fixes

Latest
Latest

Choose a tag to compare

View all tags
@Betanu701 Betanu701 released this 01 Mar 23:48
v1.0.1
9c20c6a
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

v1.0.1 — Intel Arc Build Fixes

Fixes from real-world testing on Intel Arc B580 (12GB VRAM) running on Unraid 7.1.4.

Fixes

  • Dockerfile.gpu-intel: Removed redundant intel-opencl-icd intel-level-zero-gpu level-zero apt packages (already present in intel/oneapi-runtime base image, caused unmet dependency errors)
  • Dockerfile.gpu-intel: Added pip install --upgrade pip setuptools wheel before torch install for Python 3.12 compatibility
  • requirements.txt: Bumped numpy from ==1.24.0 to >=1.26.0 (fixes pkgutil.ImpImporter removal in Python 3.12)
  • inference.py: Fixed GPU acceleration display to use is_gpu(device) instead of torch.cuda.is_available() (was showing "No" on Intel XPU)

Verified Working

  • ✅ Docker image builds successfully
  • ✅ SNAC model loads on xpu:0
  • ✅ TTS generation produces valid 24kHz WAV audio
  • ✅ device.py detects Intel XPU correctly
  • ✅ llama.cpp SYCL inference: ~20 tok/s

Docker Images

docker pull ghcr.io/betanu701/orpheus-fastapi:1.0.1-gpu-intel   # Intel Arc
docker pull ghcr.io/betanu701/orpheus-fastapi:1.0.1-gpu-cuda    # NVIDIA
docker pull ghcr.io/betanu701/orpheus-fastapi:1.0.1-gpu-rocm    # AMD
docker pull ghcr.io/betanu701/orpheus-fastapi:1.0.1-cpu          # CPU only
Assets 2
Loading