Instructions to use kernels-community/flash-attn3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Kernels
How to use kernels-community/flash-attn3 with Kernels:
# !pip install kernels from kernels import get_kernel kernel = get_kernel("kernels-community/flash-attn3") - Notebooks
- Google Colab
- Kaggle
Suport CUDA 12.8
#1
by pvalois - opened
Hello, thanks for working on this kernel!
It seems to only support CUDA 12.6 rn, is there plans to support 12.8? I get this error:
FileNotFoundError: Kernel kernels-community/flash-attn3 at revision main does not have build: torch27-cxx11-cu128-x86_64-linux
thank you for adding more versions! If possible, could you also add for aarm64? I would like to try it on a GH200 system
pvalois changed discussion status to closed