microsoft / CodeBERT

CodeBERT
MIT License
2.09k stars 430 forks source link

what if I convert the codebert to longer tokenize? #281

Closed brness closed 11 months ago

brness commented 11 months ago

In this :https://github.com/microsoft/CodeBERT/issues/278#issuecomment-1646589487, I know we can expand the length of longcoder to unlimit length, but I can also expand the codebert maximum length by this function, could this be a good choice? Do I have to re-train the model from the beginning?

guoday commented 11 months ago

Yes. It's good choice. You don't need to re-pretrain the model. Please refer to this issue

brness commented 11 months ago

Thanks for ur kindly help~~