This dataset is associated with the paper TituLLMs: A Family of Bangla LLMs with Comprehensive Benchmarking
TituLM Bangla Corpus is one of the largest Bangla clean corpus prepared for pretraining, continual pretraining or fine-tuning Large Language Model(LLM) for improving Bangla text generation capability.
This dataset contains diverse sources and categories of Bangla text. The largest part of this dataset contains filtered common crawled datasets. As we saw… See the full description on the dataset page:
https://huggingface.co/datasets/hishab/titulm-bangla-corpus.
