more

app.py

@@ -22,7 +22,7 @@ print("Model loaded successfully!")
 # Gradio Response Function
 from transformers import TextStreamer
 
-def respond(message, max_new_tokens, temperature, system_message=""):
+def respond(message, max_new_tokens, temperature, system_message="You are a helpful assistant. You should reply to the user's message without repeating the input."):
     try:
         # Prepare input messages
         messages = [{"role": "system", "content": system_message}] if system_message else []
@@ -52,10 +52,13 @@ def respond(message, max_new_tokens, temperature, system_message=""):
         # Decode the generated tokens back to text
         generated_text = tokenizer.decode(output[0], skip_special_tokens=True)
 
-        # Debug: Show the generated text
-        print("[DEBUG] Generated Text:", generated_text)
+        # Clean up the response by removing unwanted parts (e.g., system and user info)
+        cleaned_response = generated_text.split("\n")[-1]  # Assuming the response ends at the last line
 
-        return generated_text
+        # Debug: Show the cleaned response
+        print("[DEBUG] Cleaned Response:", cleaned_response)
+
+        return cleaned_response
 
     except Exception as e:
         # Debug: Log errors