Khanin Udomchoksakul
Add optimized CUDA kernels (RMSNorm, MLP, Attention) and patched Llama-3.1 modeling file
6f657bb verified | from setuptools import setup | |
| setup( | |
| name="rmsnorm_ext", | |
| packages=["rmsnorm_cuda"], | |
| package_data={"rmsnorm_cuda": ["*.so"]}, | |
| ) | |