Dear all,
I would like to compare the embedding performance between fine-tuned CodeBERT and UniXcoder. I followed the instructions on https://github.com/microsoft/CodeXGLUE/tree/main/Text-Code/NL-code-search-Adv
This page outputs MRR on fine-tuned CodeBERT but if I want to get MRR on UniXcoder, how can I modify the code?
On the other hand, if I want to get similarity between PL and NL on fine-tuned CodeBERT as described for UniXcoder in this page: https://github.com/microsoft/CodeBERT/tree/master/UniXcoder
, can I just load the fine-tuned CodeBERT with the same code?
Thank you in advance.