use chat temploate try 2

utils/models.py

@@ -160,8 +160,9 @@ def run_inference(model_name, context, question):
         if generation_interrupt.is_set():
             return ""
 
-        outputs = pipe(formatted, **generation_kwargs)
-        result = outputs[0]['generated_text'][-1]['content']
+        outputs = pipe(formatted, skip_special_tokens=True, **generation_kwargs, )
+        #print(outputs[0]['generated_text'])
+        result = outputs[0]['generated_text']
 
     except Exception as e:
         print(f"Error in inference for {model_name}: {e}")