CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-10-30 08:42:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

kv-cache : support layer reuse (#15504)

Browse Source 
* kv-cache : support layer reuse

ggml-ci

* cont : update comments [no ci]
This commit is contained in:
Georgi Gerganov
2025-08-24 13:07:07 +03:00
committed by GitHub
parent c9a24fb932
commit b730706a49
12 changed files with 203 additions and 136 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
src/llama-kv-cache.h
View File

@@ -21,9 +21,6 @@ class llama_kv_cache : public llama_memory_i {
public:
static uint32_t get_padding(const llama_cparams & cparams);
// this callback is used to filter out layers that should not be included in the cache
using layer_filter_cb = std::function<bool(int32_t il)>;
struct stream_copy_info {
bool empty() const {
assert(ssrc.size() == sdst.size());
@@ -82,18 +79,19 @@ public:
using slot_info_vec_t = std::vector<slot_info>;
llama_kv_cache(
const llama_model & model,
layer_filter_cb && filter,
ggml_type type_k,
ggml_type type_v,
bool v_trans,
bool offload,
bool unified,
uint32_t kv_size,
uint32_t n_seq_max,
uint32_t n_pad,
uint32_t n_swa,
llama_swa_type swa_type);
const llama_model & model,
ggml_type type_k,
ggml_type type_v,
bool v_trans,
bool offload,
bool unified,
uint32_t kv_size,
uint32_t n_seq_max,
uint32_t n_pad,
uint32_t n_swa,
llama_swa_type swa_type,
const layer_filter_cb & filter,
const layer_reuse_cb & reuse);
~llama_kv_cache() = default;