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

partially working, but the blas matmul is broken

Browse source
This commit is contained in:
Concedo 2023-05-13 11:35:38 +08:00
parent b335f73a60
commit 05cf5f7d6e
8 changed files with 53 additions and 21 deletions
Download patch file Download diff file Expand all files Collapse all files

2
otherarch/gpt2_v2.cpp
View file

@ -352,7 +352,7 @@ bool gpt2_eval(
if (mem_per_token > 0 && (mem_per_token*N*2 + 48u*1024*1024) > buf_size) {
const size_t buf_size_new = 320u*1024*1024 + 2*(mem_per_token*N); // add 10% to account for ggml object overhead
printf("\n%s: reallocating buffer from %zu to %zu bytes\n", __func__, buf_size, buf_size_new);
//printf("\n%s: reallocating buffer from %zu to %zu bytes\n", __func__, buf_size, buf_size_new);
// reallocate
if (buf_size_new > buf_size)