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

llama : deprecate llama_kv_self_ API (#14030)

Browse Source 
* llama : deprecate llama_kv_self_ API

ggml-ci

* llama : allow llama_memory_(nullptr)

ggml-ci

* memory : add flag for optional data clear in llama_memory_clear

ggml-ci
This commit is contained in:
Georgi Gerganov
2025-06-06 14:11:15 +03:00
committed by GitHub
parent 487a5e0401
commit 745aa5319b
34 changed files with 206 additions and 127 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
src/llama-kv-cache-unified.cpp
View File

@@ -129,13 +129,15 @@ llama_kv_cache_unified::llama_kv_cache_unified(
}
}
void llama_kv_cache_unified::clear() {
void llama_kv_cache_unified::clear(bool data) {
cells.reset();
head = 0;
for (auto & buf : bufs) {
ggml_backend_buffer_clear(buf.get(), 0);
if (data) {
for (auto & buf : bufs) {
ggml_backend_buffer_clear(buf.get(), 0);
}
}
}
@@ -1319,7 +1321,7 @@ void llama_kv_cache_unified::state_read(llama_io_read_i & io, llama_seq_id seq_i
if (!res) {
if (seq_id == -1) {
clear();
clear(true);
} else {
seq_rm(seq_id, -1, -1);
}
@@ -1500,7 +1502,7 @@ bool llama_kv_cache_unified::state_read_meta(llama_io_read_i & io, uint32_t cell
return false;
}
clear();
clear(true);
for (uint32_t i = 0; i < cell_count; ++i) {
llama_pos pos;