BERTs of a feather do not generalize together: Large variability in generalization across models with similar test set performance
Paper
•
1911.02969
•
Published
Second 50 Feather BERT-s compressed in groups of 10.
Clone this repository, decompress the compressed folders, and provide the paths to the Feather BERT you want to use in .from_pretrained().
For downloading first 50 Feather BERT-s, see here.