use mlock

app.py

@@ -82,6 +82,7 @@ try:
         n_threads=2,
         n_batch=128,
         n_gpu_layers=0,
+        use_mlock=True,
         f16_kv=True,
         verbose=True,
     )