ClaudeBot/llama.cpp
0
0
Fork 1
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-17 20:31:47 +02:00
Code Issues Packages Projects Releases Wiki Activity

server : allow model downloads at model limit fix issue #26809 (#28530)

Browse Source
This commit is contained in:
MiaoMing Chen
2026-09-12 11:50:35 +02:00
committed by GitHub
parent c8edceb061
commit 56381e407c
2 changed files with 8 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/server/server-models.cpp
+2 -1
View File
@@ -1129,7 +1129,8 @@ void server_models::load(const std::string & name, const load_options & opts) {
// exceeding models_max. Without this, the window between unload_lru() // exceeding models_max. Without this, the window between unload_lru()
// releasing its lock and this lock_guard acquiring allows multiple // releasing its lock and this lock_guard acquiring allows multiple
// threads to each observe capacity and all proceed to load. // threads to each observe capacity and all proceed to load.
if (base_params.models_max > 0) { // Download workers do not use models_max slots.
if (opts.mode == SERVER_CHILD_MODE_NORMAL && base_params.models_max > 0) {
size_t count_active = 0; size_t count_active = 0;
for (const auto & m : mapping) { for (const auto & m : mapping) {
if (m.second.meta.is_running()) { if (m.second.meta.is_running()) {
tools/server/tests/unit/test_router.py
+6 -1
View File
@@ -540,13 +540,17 @@ def _wait_for_sse_event(collected: list, event_type: str, model: str, timeout: i
def test_router_download_model(): def test_router_download_model():
"""Case 1: download a model, verify SSE events and GET /models.""" """Case 1: download a model at the model limit, verify SSE events and GET /models."""
global server global server
server.models_max = 1
server.start() server.start()
# Ensure the model is not present before we start # Ensure the model is not present before we start
server.make_request("DELETE", f"/models?model={MODEL_DOWNLOAD_ID}") server.make_request("DELETE", f"/models?model={MODEL_DOWNLOAD_ID}")
# A download worker must not consume or evict a model slot
_load_model_and_wait(MODEL_B, timeout=120)
sse_events: list = [] sse_events: list = []
stop = threading.Event() stop = threading.Event()
sse_ready = threading.Event() sse_ready = threading.Event()
@@ -580,6 +584,7 @@ def test_router_download_model():
# Model should now appear in GET /models # Model should now appear in GET /models
ids = _get_model_ids(is_reload=False) ids = _get_model_ids(is_reload=False)
assert MODEL_DOWNLOAD_ID in ids, f"{MODEL_DOWNLOAD_ID} not found in /models after download" assert MODEL_DOWNLOAD_ID in ids, f"{MODEL_DOWNLOAD_ID} not found in /models after download"
assert _get_model_status(MODEL_B) == "loaded"
def test_router_delete_model(): def test_router_delete_model():