Tokenizer used for all BLOOM models

Tokenizer information are provided at https://huggingface.co/bigscience/bloom#preprocessing

TODO: point to paper once it comes out with extra details on the tokenizer