bert-base-15lang-cased

We are sharing smaller versions of bert-base-multilingual-cased that handle a custom number of languages.

Unlike distilbert-base-multilingual-cased, our versions give exactly the same representations produced by the original model which preserves the original accuracy.

The measurements below have been computed on a Google Cloud n1-standard-1 machine (1 vCPU, 3.75 GB):

Model Num parameters Size Memory Loading time
bert-base-multilingual-cased 178 million 714 MB 1400 MB 4.2 sec
Geotrend/bert-base-15lang-cased 141 million 564 MB 1098 MB 3.1 sec

Handled languages: en, fr, es, de, zh, ar, ru, vi, el, bg, th, tr, hi, ur and sw.

For more information please visit our paper: Load What You Need: Smaller Versions of Multilingual BERT.

How to use

from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained("Geotrend/bert-base-15lang-cased")
model = AutoModel.from_pretrained("Geotrend/bert-base-15lang-cased")

To generate other smaller versions of multilingual transformers please visit our Github repo.

How to cite

@inproceedings{smallermbert,
  title={Load What You Need: Smaller Versions of Multilingual BERT},
  author={Abdaoui, Amine and Pradel, Camille and Sigel, Grégoire},
  booktitle={SustaiNLP / EMNLP},
  year={2020}
}

Contact

Please contact amine@geotrend.fr for any question, feedback or request.

New

Select AutoNLP in the “Train” menu to fine-tune this model automatically.

Downloads last month
46
Hosted inference API
Fill-Mask
Mask token: [MASK]
Examples
Examples
This model can be loaded on the Inference API on-demand.