vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 17:14:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

rephrase tensor moved warning, cleanup and prepare for ci

Browse source
This commit is contained in:
Concedo 2024-12-19 22:57:43 +08:00
parent fbf1345a66
commit 50648de0af
13 changed files with 25 additions and 2867 deletions
Download patch file Download diff file Expand all files Collapse all files

3
ggml/src/ggml-cpu/ggml-cpu-aarch64.cpp
View file

@ -4167,7 +4167,8 @@ static void flag_aarch_prepacked_quant(int type)
static const ggml::cpu::tensor_traits * ggml_aarch64_get_optimal_repack_type(const struct ggml_tensor * cur) {
if (cur->type == GGML_TYPE_Q4_0) {
if (ggml_cpu_has_avx2() || (ggml_cpu_has_sve() && ggml_cpu_has_matmul_int8() && ggml_cpu_get_sve_cnt() == QK8_0)) {
//we shall just use the regular avx2 handling, no repacking
if (/*ggml_cpu_has_avx2() ||*/ (ggml_cpu_has_sve() && ggml_cpu_has_matmul_int8() && ggml_cpu_get_sve_cnt() == QK8_0)) {
if (cur->ne[1] % 8 == 0) {
return &ggml::cpu::aarch64::q4_0_8x8_q8_0;
}