wheels.txt

# Additional Python Wheels
# 
# List wheel URLs here (one per line) for packages that need special versions
# These are installed AFTER PyTorch but BEFORE custom node dependencies
#
# The most important wheel here is Nunchaku, which provides 4-bit quantization
# support for massive speedups on Blackwell GPUs.
# You may need to change the user/__manager/config.ini setting 
# security_level=weak to get nunchaku to install, it relaxes the rule so the
# installer works.
#
# Find the latest Nunchaku releases here:
# https://github.com/nunchaku-ai/nunchaku/releases
#
# CRITICAL: The wheel filename must match your .env configuration:
# - PyTorch version (e.g., torch2.11 for PyTorch 2.11.x)
# - Python version from PYTHON_VERSION in .env (e.g., cp312 for Python 3.12)
# - Platform: linux_x86_64 (standard Linux)
#
# Current example matches:
# - PyTorch 2.11.x (torch2.11)
# - Python 3.12 (cp312)
# - Linux x86_64
#
# If you update PyTorch in .env, update this wheel URL to match!

https://github.com/nunchaku-ai/nunchaku/releases/download/v1.2.1/nunchaku-1.2.1+cu13.0torch2.10-cp312-cp312-linux_x86_64.whl

# Add other wheels below as needed
# Example: A custom build of a package
# https://example.com/path/to/custom-package-1.0.0-cp312-cp312-linux_x86_64.whl
# If you are struggling with this just ask your favourite AI to help.