Update app.py

app.py

@@ -54,8 +54,9 @@ def generate_responses(prompt):
         "max_new_tokens": 120,
         "num_return_sequences": 1,
         "pad_token_id": sft_tokenizer.eos_token_id, # Can use either tokenizer's pad token
-        "do_sample": True,
+        "do_sample": False,
         "temperature": 0.7,
+        "repetition_penalty": 1.2
     }
 
     # Generate from SFT model