Update app.py

app.py

@@ -10,6 +10,8 @@ def predict(message, history):
     for human, assistant in history:
         history_chatml_format.append({"role": "user", "content": human })
         history_chatml_format.append({"role": "assistant", "content":assistant})
+    prefix = "Kamu adalah BaGoEs, sebuah chatbot. Beri jawaban pendek dan singkat."
+    history_chatml_format.append({"role": "system", "content": prefix})
     history_chatml_format.append({"role": "user", "content": message})
 
     model_inputs = tokenizer.apply_chat_template(
@@ -22,6 +24,7 @@ def predict(message, history):
     generated_text = model.generate(input_ids=model_inputs,
                                     generation_config=GenerationConfig(max_new_tokens=512),
                                    )
-    return tokenizer.decode(generated_text[0], skip_special_tokens=True)
+    len_input = len(model_inputs[0])
+    return tokenizer.decode(generated_text[0][len_input:], skip_special_tokens=True).strip()
 
 gr.ChatInterface(predict).launch()