tune llama paramters

app.py

@@ -116,9 +116,9 @@ def try_load_model(path):
             n_ctx=2048,  # Reduced context window
             n_threads=2,
             n_threads_batch=1,
-            n_batch=64,
+            n_batch=256,
             n_gpu_layers=0,
-            use_mlock=False,
+            use_mlock=True,
             use_mmap=True,
             verbose=False,
             logits_all=True,