Train Xtransformer using PubMedBert with params inspired from paper. TBD.
Closed this issue · 0 comments
nsorros commented
The idea here is to re read the Xtransformers paper and understand which params could be switched to get a better performance based on what they did on other datasets.