- Multiprocessing during tokenization is now turned on by default. If you experience any instability, this can be turned off by setting `use_multiprocessing=False`
0.60.6
Changed
- Multiprocessing during tokenization is now turned off by default. You can enable this by setting `use_multiprocessing=True`. However, the latest Pytorch versions seems to be unstable when using multiprocessing.
0.60.3
Changed
- Multiprocessing is now turned off by default for evaluation. This is to avoid potential errors when doing evaluation during training. You can enable this by setting `use_multiprocessing_for_evaluation` to `True`.
0.60.2
0.60.1
0.60.00.60.0
Added
- Added class weights support for Longformer classification - Added new classification models: - SqueezeBert - DeBERTa - MPNet
Changed
- Updated ClassificationModel logic to make it easier to add new models