Is the extra space issue in TabbyML/CodeQwen-7B model a known flaw?

I'm using the TabbyML/CodeQwen-7B model, but the completion looks a little weird. Most of the time, it (wrongly) adds an extra space before the generated code. Not sure it's the model's issue or tabby's issue.


graceking lau

Asked on May 30, 2024

Yes, the extra space issue in the TabbyML/CodeQwen-7B model is a known flaw. Other models like deepseek do not have this issue.

May 30, 2024Edited by