]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
server : add SWA checkpoints (#15293)
authorGeorgi Gerganov <redacted>
Thu, 14 Aug 2025 11:59:50 +0000 (14:59 +0300)
committerGitHub <redacted>
Thu, 14 Aug 2025 11:59:50 +0000 (14:59 +0300)
commitd32e03f4495d3efa1c5126f53b449f1d429c5664
treef7426ecfd0fc9d9ab769ae86eeabeaf4bd545b45
parent3973163bff40f7f5161b0f08a0011729e2b0406a
server : add SWA checkpoints (#15293)

* 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
15 files changed:
common/arg.cpp
common/common.h
include/llama.h
src/llama-context.cpp
src/llama-context.h
src/llama-kv-cache-unified-iswa.cpp
src/llama-kv-cache-unified-iswa.h
src/llama-kv-cache-unified.cpp
src/llama-kv-cache-unified.h
src/llama-memory-hybrid.cpp
src/llama-memory-hybrid.h
src/llama-memory-recurrent.cpp
src/llama-memory-recurrent.h
src/llama-memory.h
tools/server/server.cpp