Meta AI issue to append duplicated messages after the answer when questions have context

[summerkiss]

Before submitting a bug, please make sure the issue hasn't been already addressed by searching through the FAQs and existing/past issues

Describe the bug

when the questions have context, the llama will return a duplicated message " I am still improving my command of other languages, and I may make errors while attempting them.".For example, if the question is the second following question is will have two of these messages after the answer.

Minimal reproducible example

<Remember to wrap the code in ```triple-quotes blocks```>

# sample code to repro the bug

questions:
大语言模型的参数是什么
分别举个例子
嵌入层的向量是什么
单词是如何拆分成向量的
transformer 结构是什么

Output

<Remember to wrap the output in ```triple-quotes blocks```>

<paste stacktrace and other outputs here>

Runtime Environment

• Model: [eg: llama-2-7b-chat]
• Using via huggingface?: [yes/no]
• OS: [eg. Linux/Ubuntu, Windows]
• GPU VRAM:
• Number of GPUs:
• GPU Make: [eg: Nvidia, AMD, Intel]

Additional context
I could not find where to report the Meta AI issue, so pasted it here.
the issue found on Aug 6