Fix: GPU duration 240s (max allowed)

app.py

@@ -48,7 +48,7 @@ print("✅ Modell geladen!")
 
 
 # ─── GPU-Funktion (nur hier ist echte GPU) ──────────────────────────────
-@spaces.GPU(duration=120)
+@spaces.GPU(duration=240)
 def generate_response(messages, temperature, max_tokens):
     """Generate a response from the model. GPU is real inside this function."""
     text = tokenizer.apply_chat_template(