ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

llama : allow custom list of swa_layers (#13726)

Browse source
This commit is contained in:
Xuan-Son Nguyen 2025-05-23 17:07:04 +02:00 • committed by GitHub
parent 9ecf3e66a3
commit 8a2afb7520
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 54 additions and 23 deletions
Download patch file Download diff file Expand all files Collapse all files

22
src/llama-hparams.cpp
View file

@ -2,6 +2,26 @@
#include "ggml.h"
llama_hparams::llama_hparams() {
swa_layers.fill(false);
}
void llama_hparams::set_swa_pattern(uint32_t n_pattern) {
for (uint32_t il = 0; il < n_layer; ++il) {
swa_layers[il] = n_pattern == 0 || (il % n_pattern < (n_pattern - 1));
}
}
bool llama_hparams::is_swa_any() const {
for (uint32_t il = 0; il < n_layer; ++il) {
if (swa_layers[il]) {
return true;
}
}
return false;
}
uint32_t llama_hparams::n_head(uint32_t il) const {
if (il < n_layer) {
return n_head_arr[il];
@ -72,7 +92,7 @@ uint32_t llama_hparams::n_embd_v_s() const {
bool llama_hparams::is_swa(uint32_t il) const {
if (il < n_layer) {
return n_swa_pattern == 0 || (il % n_swa_pattern < (n_swa_pattern - 1));
return swa_layers[il];
}
GGML_ABORT("fatal error");