Fine-tuned model sometimes repeats itself verbatim

Prompting
2

I am having a similar issue. It works when tested in playground but when tested with SDK it repeats itself by answering its own assistance response.

In my case, i am using fined fine-tuned model of gpt3.5turbo.