mlwheels 0.1.0

Auto-detect and install pre-built wheels for Flash Attention & vLLM

Pre-Built Wheels

Pre-built wheels for Flash Attention & vLLM. Skip the compilation.

Quick Install

Auto-detect your environment and install the right wheel:

pip install mlwheels

# Install Flash Attention
mlwheels flash-attn

# Install vLLM
mlwheels vllm

# Just detect environment (no install)
mlwheels --detect

Features

• Search and filter by CUDA, Python, PyTorch, and Platform
• One-click copy for pip and uv install commands
• Direct download links
• Flash Attention 2 & 3 support
• vLLM wheels for multiple CUDA versions

Supported Configurations

Flash Attention 2

• CUDA: 11.8, 12.1, 12.2, 12.3, 12.4, 12.6
• PyTorch: 2.0 - 2.10
• Python: 3.8 - 3.12
• Platforms: Linux x86_64, Linux ARM64, Windows

Flash Attention 3

• CUDA: 12.6, 12.8, 12.9, 13.0
• PyTorch: 2.8 - 2.10
• Python: 3.10 - 3.12
• Platforms: Linux x86_64, Linux ARM64, Windows

vLLM

• CUDA: 11.8, 12.1, 12.4, 12.6, 12.8, 12.9, 13.0, CPU
• Python: 3.8+
• Platforms: Linux x86_64, Linux ARM64

Sources

Flash Attention

• flashattn.dev
• Flash Attention 3 Wheels
• mjun0812/flash-attention-prebuild-wheels
• Dao-AILab/flash-attention

vLLM

• vLLM GitHub Releases
• vLLM Documentation

License

MIT