Update app.py

app.py

@@ -29,7 +29,8 @@ else:
 llm = Llama(
     model_path=model_path,
     n_ctx=1024,
-    n_threads=4  # Adjust for your CPU
+    n_threads=os.cpu_count(),  # Adjust for your CPU
+    n_batch=64
 )
 
 @app.get("/")
@@ -54,6 +55,8 @@ async def generate(request: Request):
 
     print("📤 Raw model response:", response)
 
+    llm.reset()
+
     return {
         "response": response["choices"][0]["message"]["content"].strip()
     }