lassl / lassl

Easy Language Model Pretraining leveraging Huggingface's Transformers and Datasets
Apache License 2.0
127 stars 14 forks source link

Refactor collator, tokenizer-training for ul2, t5 #104

Open wavy-jung opened 2 years ago

wavy-jung commented 2 years ago

TODO

cc: @seopbo @DaehanKim