metadata
license: mit
language:
- en
hmByT5 - Preliminary Language Models
Preliminary Historic Multilingual and Monolingual ByT5 Models. Following languages are currently covered:
- English (British Library Corpus - Books)
More details can be found in our GitHub repository.
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.
This model was trained with mean_noise_span_length=20.
Evaluation on Downstream Tasks (NER)
We evaluated the hmByT5 Base model on English AjMC dataset:
| Configuration | Run 1 | Run 2 | Run 3 | Run 4 | Run 5 | Avg. |
|---|---|---|---|---|---|---|
wsFalse-bs8-e10-lr0.00015-poolingfirst |
86.51 | 87.2 | 86.22 | 85.78 | 86.46 | 86.43 ± 0.46 |
wsFalse-bs4-e10-lr0.00016-poolingfirst |
86.12 | 87.04 | 87.01 | 85.25 | 86.74 | 86.43 ± 0.68 |
wsFalse-bs8-e10-lr0.00016-poolingfirst |
86.49 | 85.27 | 86.12 | 86.29 | 85.61 | 85.96 ± 0.45 |
wsFalse-bs4-e10-lr0.00015-poolingfirst |
86.33 | 86.05 | 84.48 | 85.68 | 86.16 | 85.74 ± 0.67 |
The ByT5 Small model achieves 85.65 ± 1.21 on this dataset.
Acknowledgements
Research supported with Cloud TPUs from Google's TPU Research Cloud (TRC). Many Thanks for providing access to the TPUs ❤️