Understanding BLOOM: An empirical study on diverse NLP tasks