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

allow quantkv with contextshift

Browse source
This commit is contained in:
Concedo 2025-03-16 21:48:42 +08:00
parent e466ce65e2
commit 6888f5495d
2 changed files with 6 additions and 7 deletions
Download patch file Download diff file Expand all files Collapse all files

4
src/llama-context.cpp
View file

@ -720,7 +720,7 @@ void llama_context::kv_self_update() {
printf("\nWARNING: The current context does not support K-shift!\n");
} else {
LLAMA_LOG_DEBUG("%s: applying K-shift\n", __func__);
// LLAMA_LOG_DEBUG("%s: applying K-shift\n", __func__);
// apply K-shift if needed
if (model.hparams.rope_type != LLAMA_ROPE_TYPE_NONE) {
@ -774,7 +774,7 @@ void llama_context::kv_self_update() {
// reserve a worst case graph if needed
if (need_reserve) {
LLAMA_LOG_DEBUG("%s: reserving a worst case graph\n", __func__);
// LLAMA_LOG_DEBUG("%s: reserving a worst case graph\n", __func__);
// build worst-case graph
uint32_t n_seqs = 1; // TODO: worst-case number of sequences