server: add dedup-cache-models preset option (#27346)

This commit is contained in:
Xuan-Son Nguyen
2026-08-19 11:04:26 +02:00
committed by GitHub
parent 98d1e92c21
commit ee4c505a4f
8 changed files with 132 additions and 2 deletions
+6
View File
@@ -4658,6 +4658,12 @@ void common_params_add_preset_options(std::vector<common_arg> & args) {
[](common_params &, int) { /* unused */ }
).set_env(COMMON_ARG_PRESET_STOP_TIMEOUT).set_preset_only());
args.push_back(common_arg(
{"dedup-cache-models"}, "0|1",
"in server router mode, hide a cached model from the model list when this preset resolves to the same model file",
[](common_params &, const std::string &) { /* unused */ }
).set_env(COMMON_ARG_PRESET_DEDUP_CACHE_MODELS).set_preset_only());
// args.push_back(common_arg(
// {"pin"},
// "in server router mode, do not unload this model if models_max is exceeded",