flash-attn-wheels / README.md
nopesadly's picture
Upload README.md with huggingface_hub
f5e7f73 verified
|
Raw
History Blame Contribute Delete
464 Bytes
# Flash-Attention Wheels for T4 GPU
Pre-compiled `.whl` files for NVIDIA T4 (sm_75).
| File | Python | PyTorch | CUDA |
|------|--------|---------|------|
| `flash_attn-2.8.3-cp312-cp312-linux_x86_64.whl` | 3.12 | 2.10.0+cu128 | 12.8 |
## Install
```bash
pip install "https://huggingface.co/nopesadly/flash-attn-wheels/resolve/main/flash_attn-2.8.3-cp312-cp312-linux_x86_64.whl"
```
> Compiled with `TORCH_CUDA_ARCH_LIST=7.5` — only compatible with T4 GPUs.