Fine tuning of various BERT-like transformer models

Description: Various pre-trained language models are fine-tuned and tested on publicly available datasets.

Contents

• Fine tuning of Hindi-BERT with Keras using a review dataset

• Fine tuning of T-XLM-RoBERTa-base model using transformers Trainer class on UMSAB Hindi sentiment analysis dataset

• A general fine-tuning recipe of (BERT like) Hugging Face transformer models using native PyTorch (training workflow is demonstrated on German UMSAB dataset)

• A general hyperparameter optimization recipe for BERT-like Hugging Face transformer models by using the transformers Trainer class (on UMSAB dataset).