/TransformersClassification

Primary LanguagePython

TransformersClassification

This package implements different methods of classification using Transformers. It intends to demonstrate the capabilities or pre-trained BERT transformer, with no additional re-training, versus fine-tuning on a small set of samples.

  1. Based on Next Sentence Prediction, over a pre-trained vanilla BERT See: TransformersClassification/src/classifiers/category_prediction_next_sentence.py
  2. Based on Masked Word Prediction, over a pre-trained vanilla BERT See: TransformersClassification/src/classifiers/category_prediction_mask.py
  3. Fine-tuned on downstream task. See: TransformersClassification/src/classifiers/category_classifier_prediction.py See: TransformersClassification/src/classifiers/category_classifier_test.py See: TransformersClassification/src/classifiers/category_classifier_train.py

In addition, we also include a proof of concept for semantic sentence similarity based on: See: TransformersClassification/src/classifiers/category_prediction_mask.py