Extending NERDA Library for Continual Learning

pip install NERDA-Con

Due to a recent issue with scikit-learn installation, which is an ongoing issue with the original NERDA, we have fixed it on the GitHub repository, and you can install the updated version as follows:

pip install nerda-con@git+https://github.com/SupritiVijay/nerda-con

model.train_next_task(training,validation)

The training and validation must be NERDA-optimized dataloaders.

model.evaluate_performance(test)

To set shared model parameters,

model.shared_model = model.transformer_model

@inproceedings{nerda-con,
  title = {NERDA-Con},
  author = {Supriti Vijay, Aman Priyanshu},
  year = {2022},
  publisher = {{GitHub}},
  url = {https://github.com/SupritiVijay/NERDA-Con}
}

Nerda is a framework for fine-tuning pretrained transformers for Named-Entity Recognition (NER) tasks.

@inproceedings{nerda,
  title = {NERDA},
  author = {Kjeldgaard, Lars and Nielsen, Lukas},
  year = {2021},
  publisher = {{GitHub}},
  url = {https://github.com/ebanalyse/NERDA}
}