aemami1/adept

Availability of finetuned models

Closed this issue · 3 comments

Hi there,

Thanks for putting together this repo! We would love to be able to compare to and build on your excellent work. Is it possible to release your finetuned models (say, with huggingface model sharing)? In particular, I'm interested in the BERT/RoBERTa/DeBERTa models under the 3-class/with context setting.

There are a bunch of hyperparameters mentioned in the paper but no optimal configuration is reported. It would be much easier to reproduce your results if the models themselves can be shared. Thanks in advance!

NVM, we were able to reproduce the numbers with a slightly different set of parameters. Thanks anyway!

Thank you! No worries!