char-fork/llama.cpp

git clone https://git.t4t.associates/char-fork/llama.cpp

Johannes GäßlerCUDA: faster tile FA, add oob checks, more HSs (#16492)11f0af550

char/aliceai-t5-moe
146 B5 linesraw
1// This file has been autogenerated by generate_cu_files.py, do not edit manually.
2
3#include "../fattn-tile.cuh"
4
5DECL_FATTN_TILE_CASE(128, 128);