CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-16 11:27:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

kv-cache : remove llama_kv_cache_i

Browse Source
This commit is contained in:
Georgi Gerganov
2025-02-19 14:36:27 +02:00
parent f5cedbcaaa
commit 5f11a5502a
5 changed files with 330 additions and 339 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
src/llama-kv-cache.h
View File

@@ -49,7 +49,7 @@ struct llama_kv_cache_slot_info {
// TODO: pimpl
// TODO: add notion of max sequences
// TODO: add llama_hparams &
struct llama_kv_cache : public llama_graph_kv_cache_i {
struct llama_kv_cache {
llama_kv_cache(const llama_hparams & hparams);
virtual ~llama_kv_cache() = default;
@@ -97,19 +97,6 @@ struct llama_kv_cache : public llama_graph_kv_cache_i {
size_t size_k_bytes() const;
size_t size_v_bytes() const;
// graph build API
virtual void build_shift(
ggml_context * ctx0,
ggml_cgraph * gf,
llama_graph_i * lgf) override;
virtual void build_defrag(
ggml_context * ctx0,
ggml_cgraph * gf,
int32_t max_nodes,
bool v_trans) override;
// state save/load
void state_write(llama_io_write_i & io, llama_seq_id seq_id = -1) const;