flash-attention/csrc/layer_norm/README.md
2022-11-13 22:30:23 -08:00

7 lines
308 B
Markdown

This CUDA extension implements fused dropout + residual + LayerNorm, based on
Apex's [FastLayerNorm](https://github.com/NVIDIA/apex/tree/master/apex/contrib/layer_norm).
We add dropout and residual, and make it work for both pre-norm and post-norm architecture.
```sh
cd csrc/layer_norm && pip install .
```