In this experiment we sample 4B bytes (~4GB of text) from each corpora (and upsample Swedish and Finnish) and train for another epoch (2 epochs in total).

Pretraining

We use the official JAX/FLAX example in Hugging Face Transformers to pretrain a ByT5 model on a single v3-8 TPU. Details about the training can be found here.

Evaluation on Downstream Tasks (NER)

We evaluated the hmByT5 model on downstream tasks:

Model	English AjMC	German AjMC	French AjMC	Finnish NewsEye	Swedish NewsEye	Dutch ICDAR	French ICDAR	Avg.
`hmbyt5-preliminary/byt5-small-multilingual-4g-2e`	83.86 ± 0.61	87.54 ± 0.19	84.29 ± 0.41

Acknowledgements

Research supported with Cloud TPUs from Google's TPU Research Cloud (TRC). Many Thanks for providing access to the TPUs ❤️

hmbyt5-preliminary/byt5-small-multilingual-4g-2e

hmByT5 - Preliminary Language Models

Pretraining

Evaluation on Downstream Tasks (NER)

Acknowledgements

hmbyt5-preliminary/
byt5-small-multilingual-4g-2e