vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

CUDA: faster large batch FA without tensor cores (#7314)

Browse source
This commit is contained in:
Johannes Gäßler 2024-05-17 18:54:52 +02:00 • committed by GitHub
parent 82ca83db3c
commit 0fc1e820a9
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
7 changed files with 823 additions and 15 deletions
Download patch file Download diff file Expand all files Collapse all files

3
ggml-cuda/fattn-tile-f16.cuh Normal file
View file

@ -0,0 +1,3 @@
#include "common.cuh"
void ggml_cuda_flash_attn_ext_tile_f16(ggml_backend_cuda_context & ctx, ggml_tensor * dst);