@hooman650 any update on how did this project go ? i tired joining via the discord link, but the channel doesnt exist now. actually i needed to refer on how you guys managed to use the seqio library to mix datasets in huggingface. currently i am struggling to use seqio in huggingface. FYI i am planning to pretrain MT5 using huggingface trainer, and using seqio mixture would be useful for multi-lingual dataset.