char-fork/llama.cpp

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

Anav PrasadCUDA: Add Flash Attention Support for Head Dimension 512 (#20998)88458164c

char/aliceai-t5-moe
439 B12 linesraw
1// This file has been autogenerated by generate_cu_files.py, do not edit manually.
2
3#include "../fattn-mma-f16.cuh"
4
5DECL_FATTN_MMA_F16_CASE(64, 64, 4, 4);
6DECL_FATTN_MMA_F16_CASE(80, 80, 4, 4);
7DECL_FATTN_MMA_F16_CASE(96, 96, 4, 4);
8DECL_FATTN_MMA_F16_CASE(112, 112, 4, 4);
9DECL_FATTN_MMA_F16_CASE(128, 128, 4, 4);
10DECL_FATTN_MMA_F16_CASE(256, 256, 4, 4);
11DECL_FATTN_MMA_F16_CASE(512, 512, 4, 4);
12DECL_FATTN_MMA_F16_CASE(576, 512, 4, 4);