To run the script, take the take the script corresponding to the model you want to use. Just run the required cells for training and inference purpose. Directly running from start to end will also train the model will also perform inference.
Below list shows the script name and the corresponding model used:-
- analysis_and_preprocessing.ipynb - This script shows few analysis done on the dataset.
- transformers-ire_bert_0.59903.ipynb - This script uses BERT model to fine-tune the dataset.
- transformers-ire_scibert_0.58162.ipynb - This script uses SciBERT model to fine-tune the dataset.
- transformers-ire_distilbert_0.55423.ipynb - This script uses DistilBERT model to fine-tune the dataset.
- ULMfit_0.41223.ipynb - This script uses ULMfit model to fine-tune the dataset.
- transformers-ire_bert_full_context.ipynb - This script uses BERT model to fine-tune model on the full context dataset.
- SBERT/ - This directory contains the code to fine-tune SBERT on the dataset.
- error_analysis.ipynb - This script contains the code used to perform the error analysis on valdiation data.
- submission_files/ - This directory contains the subsmission files submitted on kaggle to report the F1-macro score on test dataset.
F1-macro score - 0.41223
ULMfit model link - Link to model
F1-macro score - 0.59903
BERT model link - Link to model
F1-macro score - 0.58162
SciBERT model link - Link to model
F1-macro score - 0.54610
F1-macro score - 0.52670
F1-macro score - 0.55243