Update app.py

app.py

@@ -7,6 +7,7 @@ from llama_cpp import Llama
 import requests
 import tempfile
 import json
+import gc
 from concurrent.futures import ThreadPoolExecutor
 
 app = Flask(__name__)
@@ -53,7 +54,10 @@ class LLMManager:
                     use_mlock=True,
                     mmap=True,
                     low_vram=False,
-                    vocab_only=False
+                    vocab_only=False,
+                    verbose=False,
+                    logits_all=False,
+                    mul_mat_q=True
                 )
                 
                 os.remove(temp_path)
@@ -140,6 +144,7 @@ class LLMManager:
             
         finally:
             self.generation_lock.release()
+            gc.collect()
     
     def get_loaded_models(self):
         """Obtener lista de modelos cargados"""