Fine-tuned model output crazy

Hello everyone, recently I have been fine-tuning the gpt-3.5-turbo-0613 model. However, the fine-tuned model keeps generating an overwhelming number of <div> HTML tags in its output. I’m not sure whether it’s an issue with GPT itself. The attached image shows the output when I posed a question to the fine-tuned model, and it continues to output <div> tags until it exceeds the output limit.