Improper input format for chat models

[chujiezheng]

In this line, when you transform the messages into a textual prompt, you may need to add the arg add_generation_prompt=True

MathCheck/scripts/text_vllm_inference.py

Line 96 in fc0454e

prompt = tokenizer.apply_chat_template(messages, tokenize = False)

[zhouzihao501]

Hi chujie,

Thanks for your notice. We will add it later.