happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

add back ggml_cont for Vcur

This commit is contained in:
Yee Man Chan 2026-02-03 18:42:17 +08:00
parent efaea45cb3
commit 000fded1ea
1 changed files with 1 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
src/models/kimi-linear.cpp
View File

@ -291,6 +291,7 @@ llm_build_kimi_linear::llm_build_kimi_linear(const llama_model & model, const ll
ggml_row_size(kv->type, kv_per_head),
ggml_row_size(kv->type, kv_per_head * n_head),
ggml_row_size(kv->type, n_embd_head_qk_nope));
Vcur = ggml_cont(ctx0, Vcur);
cb(Vcur, "mla_V", il);
// Concatenate k_nope + k_pe (broadcast k_pe to all heads)