# SageAttention 2.2.0 — Pre-built Wheel

Pre-compiled SageAttention 2.2.0 wheel for **Linux x86_64** with CUDA 13 and C++20.

## Compatibility

| Component | Version |
|-----------|---------|
| Python | 3.12 |
| PyTorch | 2.15.0 nightly (cu132) |
| CUDA Toolkit | 13.3 |
| GPU | Ada Lovelace (sm_89) — RTX 4050/4060/4070/4080/4090 |
| SageAttention | 2.2.0 |

## Install

```bash
pip install sageattention-2.2.0-cp312-cp312-linux_x86_64.whl --no-deps
```

## Build from Source (if needed)

### Requirements
- CUDA Toolkit (nvcc)
- PyTorch with CUDA support
- 16GB+ RAM or adequate swap

### Steps

```bash
git clone --depth 1 https://github.com/thu-ml/SageAttention.git /tmp/SageAttention
cd /tmp/SageAttention

# Patch: PyTorch 2.15 nightly requires C++20
sed -i 's/-std=c++17/-std=c++20/g' setup.py

# Build
CUDA_HOME=/usr/local/cuda-13.3 \
TORCH_CUDA_ARCH_LIST="8.9" \
MAX_JOBS=1 \
NVCC_APPEND_FLAGS="--threads 8" \
python setup.py install
```

### Create Wheel

```bash
python setup.py bdist_wheel
cp dist/*.whl /path/to/wheels/
```

## Notes

- Built on Ubuntu 24.04 with 14GB RAM + 19GB swap
- `MAX_JOBS=1` used to avoid OOM during CUDA kernel compilation
- PyTorch 2.15 nightly requires C++20; upstream SageAttention ships with C++17 which must be patched
- Wheel includes sm_80 (Ampere) and sm_89 (Ada) kernels