wildminder/AI-windows-whl

Pre-compiled Python whl for Flash-attention, SageAttention, NATTEN, xFormer etc

What it solves

Building complex AI/ML Python packages from source on Windows is often a complicated and error-prone process. Many high-performance libraries, such as flash-attention and xformers, lack official pre-built wheels for the Windows platform, creating a significant barrier for developers and AI enthusiasts.

How it works

This project provides a centralized, curated collection of direct links to pre-compiled .whl (wheel) files. These wheels are tailored for specific combinations of Python, PyTorch, and CUDA versions. Users can install these libraries using pip install with the direct URL to the compatible wheel file, bypassing the need for local compilation.

Who it’s for

AI developers, enthusiasts, and specifically the ComfyUI community and other PyTorch users on Windows who need high-performance AI libraries without the struggle of building from source.

Highlights

  • Pre-compiled Wheels: Eliminates the need for complex local compilation of AI libraries.
  • Curated Collection: Provides a tailored list of compatible versions for PyTorch, Flash Attention, xformers, SageAttention, NATTEN, Triton, and bitsandbytes.
  • Searchable Interface: Includes a dedicated web page ("Find Windows AI Wheels") for quickly locating the required package version.
  • Wide Version Support: Tracks wheels for a variety of Python (3.9 - 3.14) and CUDA versions.

Related

  • Project
    opencv/opencv-python

    Pre-built Python bindings for OpenCV that allow users to install the computer vision library via pip without needing to manual compilation.

  • Project
    mjun0812/flash-attention-prebuild-wheels

    Provides pre‑built binary wheels for FlashAttention (fast attention kernels) across many Python, CUDA/ROCm, PyTorch, and OS versions, letting users install with a single pip command instead of compiling from source.

  • Project
    YanWenKun/ComfyUI-Windows-Portable

    A ready‑to‑run Windows package that bundles ComfyUI (a node‑based UI for Stable Diffusion) with 40+ custom nodes and 300+ pre‑compiled Python packages, letting users launch AI image/video generation instantly on an NVIDIA GPU.

  • Project
    scikit-learn/scikit-learn

    A Python module for machine learning built on top of SciPy, providing a wide range of tools and algorithms for ML practitioners.

  • Project
    Esri/deep-learning-frameworks

    A collection of installers that bundle 254 deep learning and machine learning libraries for use within ArcGIS Pro and Server to solve geospatial AI problems.