tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-06-29 12:35:16 +00:00
Code Issues Packages Projects Releases Wiki Activity

llama : fix Gemma3 SWA KV cache shift (#12373)

Browse Source 
* llama : fix Gemma3 SWA KV cache shift

ggml-ci

* hparams : add comment [no ci]
This commit is contained in:
Georgi Gerganov
2025-03-13 19:08:07 +02:00
committed by GitHub
parent be7c303410
commit 84d5475541
6 changed files with 37 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
src/llama-hparams.cpp
View File

@ -69,3 +69,11 @@ uint32_t llama_hparams::n_embd_v_s() const {
// corresponds to Mamba's ssm_states size
return ssm_d_state * ssm_d_inner;
}
bool llama_hparams::is_sliding(uint32_t il) const {
if (il < n_layer) {
return n_swa > 0 && n_swa_pattern > 0 && il % n_swa_pattern < (n_swa_pattern - 1);
}
GGML_ABORT("fatal error");
}