I have enabled CACHE=1, there is a long wait in the middle of the conversation

[abc20220327]

I have enabled CACHE=1, the API I use, in the process of continuous dialogue, every time I ask a question, I need to wait a lot of time, whether the model will be reloaded every time the complete call is made, why is it so slow?

[HashemAlsaket]

@abc20220327 would you please provide the code to reproduce the error?