JusteLeo/Nunchaku-Zimage-Win-Wheels

Nunchaku Prebuilt Wheels for Z-image (Windows)

This repository provides pre-compiled binaries (.whl) for the Nunchaku library. These builds include Z-image support and are specifically optimized for Windows x64 environments. The latest wheels now support:

• RTX 3000 Series (Ampere)
• RTX 4000 Series (Ada Lovelace)
• Data Center GPUs (A100, H100)

---

Compatibility Matrix (Windows Only)

All wheels target CUDA 12.8.

Python Version	Torch 2.7.0 (cu128)	Torch 2.8.0 (cu128)
Python 3.11	✅ Available	✅ Available
Python 3.12	✅ Available	✅ Available
Python 3.13	✅ Available	✅ Available

---

Build Methodology

Built in December 2025 with the following configuration:

• OS: Windows 11 x64
• Compiler: MSVC (Visual Studio 2022) + NVCC 12.8
• CUDA: v12.8

Build Commands (For local compilation)

If you wish to build the library yourself, the script will automatically detect and optimize for your specific GPU architecture:

set CUDA_HOME=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.8
set CUDA_PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.8
set DISTUTILS_USE_SDK=1

uv pip install torch==2.x.0 torchvision --index-url https://download.pytorch.org/whl/cu128
uv pip install numpy ninja setuptools packaging wheel
uv build --wheel --no-build-isolation

---

Installation

1. Download the .whl file matching your Python and Torch version from the Files and versions tab.
2. Install via pip:

# Example for Python 3.12 and Torch 2.8.0
pip install nunchaku-1.1.0+torch2.8-cp312-cp312-win_amd64.whl

3. Verify the installation:

import nunchaku
print("Success: Nunchaku (Z-image) loaded.")

---

Disclaimer

This is an unofficial community repository for Windows users. For the original source code and official updates, please visit the Nunchaku-tech GitHub.