tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-29 04:35:05 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : support GEGLU for jina-bert-v2 (#14090)

Browse Source
This commit is contained in:
Sigbjørn Skjæret
2025-06-10 18:02:08 +02:00
committed by GitHub
parent 652b70e667
commit 3678b838bb
4 changed files with 9 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
src/llama-graph.cpp
View File

@ -650,6 +650,7 @@ ggml_tensor * llm_graph_context::build_ffn(
{
// Project to 4h. If using swiglu double the output width, see https://arxiv.org/pdf/2002.05202.pdf
int64_t split_point = cur->ne[0] / 2;
// TODO: these conts should not be needed, see https://github.com/ggml-org/llama.cpp/pull/14090#discussion_r2137437217
ggml_tensor * x0 = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, split_point, cur->ne[1], cur->nb[1], 0));
ggml_tensor * x1 = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, split_point, cur->ne[1], cur->nb[1], split_point * ggml_element_size(cur)));
@ -663,7 +664,7 @@ ggml_tensor * llm_graph_context::build_ffn(
{
// Split into two equal parts
int64_t split_point = cur->ne[0] / 2;
// TODO: these conts should not be needed
// TODO: these conts should not be needed, see https://github.com/ggml-org/llama.cpp/pull/14090#discussion_r2137437217
ggml_tensor * x0 = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, split_point, cur->ne[1], cur->nb[1], 0));
ggml_tensor * x1 = ggml_cont(ctx0, ggml_view_2d(ctx0, cur, split_point, cur->ne[1], cur->nb[1], split_point * ggml_element_size(cur)));