Update app.py

app.py

@@ -24,6 +24,9 @@ try:
         model=model,
         tokenizer=tokenizer,
         torch_dtype=torch.float16,
+        max_length=128,          # Limit response length
+        num_return_sequences=1,  # Only generate one response
+        do_sample=False         # Use greedy decoding for speed
     )
 
 except Exception as e: