Public/koboldcpp
1
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-09-19 09:15:18 +02:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'master' into concedo_experimental

Browse Source 
# Conflicts:
#	CMakeLists.txt
#	Makefile
This commit is contained in:
Concedo
2023-04-30 10:35:02 +08:00
parent f149114395 c3ca7a5f05
commit 0061b90ec6
5 changed files with 114 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files
llama.cpp
+2 -3
View File
@@ -5,7 +5,7 @@
#include <cstdio>
#endif
#include "llama_util.h"
#include "llama-util.h"
#include "llama.h"
#include "ggml.h"
@@ -33,7 +33,6 @@
#define LLAMA_USE_SCRATCH
#define LLAMA_MAX_SCRATCH_BUFFERS 16
// available llama models
enum e_model {
MODEL_UNKNOWN,
@@ -790,7 +789,7 @@ static bool kv_cache_init(
const int n_embd = hparams.n_embd;
const int n_layer = hparams.n_layer;
const int64_t n_mem = (int64_t)n_layer*n_ctx;
const int64_t n_mem = n_layer*n_ctx;
const int64_t n_elements = n_embd*n_mem;
cache.buf.resize(2u*n_elements*ggml_type_size(wtype) + 2u*MB);