Edit model card

BERT-BASE-MONGOLIAN-CASED

Link to Official Mongolian-BERT repo

Model description

This repository contains pre-trained Mongolian BERT models trained by tugstugi, enod and sharavsambuu. Special thanks to nabar who provided 5x TPUs.

This repository is based on the following open source projects: google-research/bert, huggingface/pytorch-pretrained-BERT and yoheikikuta/bert-japanese.

How to use

from transformers import pipeline, AutoTokenizer, AutoModelForMaskedLM

tokenizer = AutoTokenizer.from_pretrained('tugstugi/bert-base-mongolian-cased', use_fast=False)
model = AutoModelForMaskedLM.from_pretrained('tugstugi/bert-base-mongolian-cased')

## declare task ##
pipe = pipeline(task="fill-mask", model=model, tokenizer=tokenizer)

## example ##
input_  = '[MASK] хот Монгол улсын нийслэл.'

output_ = pipe(input_)
for i in range(len(output_)):
    print(output_[i])

## output ##
# {'sequence': 'Улаанбаатар хот Монгол улсын нийслэл.', 'score': 0.826970100402832, 'token': 281, 'token_str': 'Улаанбаатар'}
# {'sequence': 'Нийслэл хот Монгол улсын нийслэл.', 'score': 0.06551621109247208, 'token': 4059, 'token_str': 'Нийслэл'}
# {'sequence': 'Эрдэнэт хот Монгол улсын нийслэл.', 'score': 0.0264141745865345, 'token': 2229, 'token_str': 'Эрдэнэт'}
# {'sequence': 'Дархан хот Монгол улсын нийслэл.', 'score': 0.017083868384361267, 'token': 1646, 'token_str': 'Дархан'}
# {'sequence': 'УБ хот Монгол улсын нийслэл.', 'score': 0.010854342952370644, 'token': 7389, 'token_str': 'УБ'}

Training data

Mongolian Wikipedia and the 700 million word Mongolian news data set [Pretraining Procedure]

BibTeX entry and citation info

@misc{mongolian-bert,
  author = {Tuguldur, Erdene-Ochir and Gunchinish, Sharavsambuu and Bataa, Enkhbold},
  title = {BERT Pretrained Models on Mongolian Datasets},
  year = {2019},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/tugstugi/mongolian-bert/}}
}
Downloads last month
80
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.