Developing the DistilBERT model with Sequence Classification on 9 newsgroup datasets for the Nepali language with 95.475% accuracy. It was trained using Transfer Learning on a pretrained model from HuggingFace called suyogyaart/nepali-newsgroup-classification.
With the help of this codebase, I managed to win IT MEET 2022's AI Compitetion Text Challenge organized by Kathmandu University and F1Soft.
The entire training and testing dataset can be found here.