CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-19 11:57:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: faster q2_K, q3_K MMQ + int8 tensor cores (#7921)

Browse Source 
* CUDA: faster q2_K, q3_K MMQ + int8 tensor cores

* try CI fix

* try CI fix

* try CI fix

* fix data race

* rever q2_K precision related changes
This commit is contained in:
Johannes Gäßler
2024-06-14 18:41:49 +02:00
committed by GitHub
parent 66ef1ceedf
commit 76d66ee0be
6 changed files with 468 additions and 330 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
ggml-cuda/softmax.cu
View File

@@ -130,6 +130,7 @@ static void soft_max_f32_cuda(const float * x, const T * mask, float * dst, cons
const float m0 = powf(2.0f, -(max_bias ) / n_head_log2);
const float m1 = powf(2.0f, -(max_bias / 2.0f) / n_head_log2);
// FIXME: this limit could be raised by ~2-4x on Ampere or newer
if (shmem < ggml_cuda_info().devices[ggml_cuda_get_device()].smpb) {
switch (ncols_x) {
case 32: