Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2025-02-02 15:02:47 +01:00
Code Issues Packages Projects Releases Wiki Activity

metal : use mm kernels for batch size > 2

Browse Source
This commit is contained in:
Georgi Gerganov 2023-09-28 16:02:20 +03:00
parent e9463792d3
commit 4c72ab13b2
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ggml-metal.m
View File

@ -958,7 +958,7 @@ void ggml_metal_graph_compute(
src1t == GGML_TYPE_F32 &&
[ctx->device supportsFamily:MTLGPUFamilyApple7] &&
ne00%32 == 0 &&
ne11 > 1) {
ne11 > 2) {
switch (src0->type) {
case GGML_TYPE_F32: [encoder setComputePipelineState:ctx->pipeline_mul_mm_f32_f32]; break;
case GGML_TYPE_F16: [encoder setComputePipelineState:ctx->pipeline_mul_mm_f16_f32]; break;