CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-05 09:36:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

server : add SWA checkpoints (#15293)

Browse Source 
* server : add SWA checkpoints

ggml-ci

* cont : server clean-up

* server : handle state restore fails

* llama : add extended llama_state_seq_ API

* server : do not make checkpoints if --swa-full

ggml-ci

* llama : remove flags value for NONE

* server : configure number of SWA checkpoints with CLI arg

ggml-ci

* args : fix scope of new argument
This commit is contained in:
Georgi Gerganov
2025-08-14 14:59:50 +03:00
committed by GitHub
parent 3973163bff
commit d32e03f449
15 changed files with 206 additions and 54 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
src/llama-memory-recurrent.cpp
View File

@@ -680,7 +680,9 @@ size_t llama_memory_recurrent::size_s_bytes() const {
return size_s_bytes;
}
void llama_memory_recurrent::state_write(llama_io_write_i & io, llama_seq_id seq_id) const {
void llama_memory_recurrent::state_write(llama_io_write_i & io, llama_seq_id seq_id, llama_state_seq_flags flags) const {
GGML_UNUSED(flags);
std::vector<std::pair<uint32_t, uint32_t>> cell_ranges; // ranges, from inclusive, to exclusive
uint32_t cell_count = 0;
@@ -718,7 +720,9 @@ void llama_memory_recurrent::state_write(llama_io_write_i & io, llama_seq_id seq
state_write_data(io, cell_ranges);
}
void llama_memory_recurrent::state_read(llama_io_read_i & io, llama_seq_id seq_id) {
void llama_memory_recurrent::state_read(llama_io_read_i & io, llama_seq_id seq_id, llama_state_seq_flags flags) {
GGML_UNUSED(flags);
uint32_t cell_count;
io.read_to(&cell_count, sizeof(cell_count));