Update app.py

app.py

@@ -377,7 +377,7 @@ def generate_response(user_input, model_id):
         formatted_prompt = "\n".join([f"{m['role']}: {m['content']}" for m in func_caller])
 
         #prompt = user_input
-        device = accelerator.device  # Automatically uses GPU or CPU based on accelerator setup
+        #device = accelerator.device  # Automatically uses GPU or CPU based on accelerator setup
         
         generation_config = GenerationConfig(
             do_sample=(highest_label == "dialog continuation" or highest_label == "recommendation request"),  # True if dialog continuation, else False
@@ -398,9 +398,9 @@ def generate_response(user_input, model_id):
         gpt_output = model.generate(gpt_inputs["input_ids"], max_new_tokens=50, generation_config=generation_config)
         final_response = tokenizer.decode(gpt_output[0], skip_special_tokens=True)
         # Extract AI's response only (omit the prompt)
-        ai_response2 = final_response.replace(reformulated_prompt, "").strip()
+        #ai_response2 = final_response.replace(reformulated_prompt, "").strip()
         ai_response = re.sub(re.escape(formatted_prompt), "", final_response, flags=re.IGNORECASE).strip()
-        ai_response = re.split(r'(?<=\w[.!?]) +', ai_response)
+        #ai_response = re.split(r'(?<=\w[.!?]) +', ai_response)
         ai_response = [s.strip() for s in re.split(r'(?<=\w[.!?]) +', ai_response) if s]
        
         # Encode the prompt and candidates