CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-07 09:57:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

improved memory management fixes

Browse Source
This commit is contained in:
slaren
2023-07-21 12:41:46 +02:00
parent 56e9ae062c
commit 3d679827e7
4 changed files with 72 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
llama.cpp
View File

@@ -703,7 +703,9 @@ static bool kv_cache_init(
const int64_t n_mem = n_layer*n_ctx;
const int64_t n_elements = n_embd*n_mem;
size_t size = 2u*n_elements*ggml_type_size(wtype) + 2u*MB;
size_t size = 2u*n_elements*ggml_type_size(wtype);
fprintf(stderr, "%s: allocating %.2f MB for kv cache\n", __func__, size / 1024.0 / 1024.0);
cache.buf = ggml_buffer_alloc(backend, size, 2);
cache.n = 0;