I read the relevant paper, I would like to finetune the base version. I am interested in the STS task. I assume the mean encoder would work best.
I would appreciate it if one can link me to the code or give me some pointers.
Moreover, can I confirm that the models on the page are the fine-tuned ones?
If so are they on the combined NLI + CommonQA?
Many thanks,
Phil