]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commitdiff
server : pass default --keep argument (#14120)
authorTaylor <redacted>
Wed, 11 Jun 2025 10:43:43 +0000 (06:43 -0400)
committerGitHub <redacted>
Wed, 11 Jun 2025 10:43:43 +0000 (13:43 +0300)
tools/server/server.cpp

index 3b5e03528e2d7b08c32b64362435d1f49d5fdd59..1b1cf439baa5fbd125dd4a5addf8d38250209cbf 100644 (file)
@@ -233,6 +233,7 @@ struct server_task {
         slot_params defaults;
         defaults.sampling    = params_base.sampling;
         defaults.speculative = params_base.speculative;
+        defaults.n_keep      = params_base.n_keep;
 
         // enabling this will output extra debug information in the HTTP responses from the server
         params.verbose           = params_base.verbosity > 9;
@@ -2060,6 +2061,7 @@ struct server_context {
             SLT_INF(slot, "new slot n_ctx_slot = %d\n", slot.n_ctx);
 
             slot.params.sampling = params_base.sampling;
+            slot.params.n_keep = params_base.n_keep;
 
             slot.callback_on_release = [this](int) {
                 queue_tasks.pop_deferred_task();