language: Chinese
datasets: CLUECorpusSmall
widget:
- text: 中国的首都是[MASK]
Chinese word-based RoBERTa Miniatures
Model description
This is the set of 5 Chinese word-based RoBERTa models pre-trained by UER-py.
Most Chinese pre-trained weights are based on Chinese character. Compared with character-based models, word-based models are faster (because of shorter sequence length) and have better performance according to our experimental results. To this end, we released the 5 Chinese word-based RoBERTa models of different sizes. In order to facilitate users to reproduce the results, we used the publicly available corpus and word segmentation tool, and provided all training details.
You can download the 5 Chinese RoBERTa miniatures either from the UER-py Github page, or via HuggingFace from the links below:
| Link | |
|---|---|
| Tiny | 2/128 (Tiny) |
| Mini | 4/256 (Mini) |
| Small | 4/512 (Small) |
| Medium | 8/512 (Medium) |
| Base | 12/768 (Base) |
How to use
You can use this model directly with a pipeline for masked language modeling (take the case of word-based RoBERTa-Medium):
>>> from transformers import pipeline
>>> unmasker = pipeline('fill-mask', model='uer/roberta-medium-word-chinese-cluecorpussmall')
>>> unmasker("[MASK]的首都是北京。")
[
{'sequence': '中国 的首都是北京。',
'score': 0.21525809168815613,
'token': 2873,
'token_str': '中国'},
{'sequence': '北京 的首都是北京。',
'score': 0.15194718539714813,
'token': 9502,
'token_str': '北京'},
{'sequence': '我们 的首都是北京。',
'score': 0.08854265511035919,
'token': 4215,
'token_str': '我们'},
{'sequence': '美国 的首都是北京。',
'score': 0.06808705627918243,
'token': 7810,
'token_str': '美国'},
{'sequence': '日本 的首都是北京。',
'score': 0.06071401759982109,
'token': 7788,
'token_str': '日本'}
]
Here is how to use this model to get the features of a given text in PyTorch:
from transformers import AlbertTokenizer, BertModel
tokenizer = AlbertTokenizer.from_pretrained('uer/roberta-medium-word-chinese-cluecorpussmall')
model = BertModel.from_pretrained("uer/roberta-medium-word-chinese-cluecorpussmall")
text = "用你喜欢的任何文本替换我。"
encoded_input = tokenizer(text, return_tensors='pt')
output = model(**encoded_input)
and in TensorFlow:
from transformers import AlbertTokenizer, TFBertModel
tokenizer = AlbertTokenizer.from_pretrained('uer/roberta-medium-word-chinese-cluecorpussmall')
model = TFBertModel.from_pretrained("uer/roberta-medium-word-chinese-cluecorpussmall")
text = "用你喜欢的任何文本替换我。"
encoded_input = tokenizer(text, return_tensors='tf')
output = model(encoded_input)
Since BertTokenizer does not support sentencepiece, AlbertTokenizer is used here.
Training data
CLUECorpusSmall is used as training data. Google's sentencepiece is used for word segmentation. The sentencepiece model is trained on CLUECorpusSmall corpus:
>>> import sentencepiece as spm
>>> spm.SentencePieceTrainer.train(input='cluecorpussmall.txt',
model_prefix='cluecorpussmall_spm',
vocab_size=100000,
max_sentence_length=1024,
max_sentencepiece_length=6,
user_defined_symbols=['[MASK]','[unused1]','[unused2]',
'[unused3]','[unused4]','[unused5]','[unused6]',
'[unused7]','[unused8]','[unused9]','[unused10]'],
pad_id=0,
pad_piece='[PAD]',
unk_id=1,
unk_piece='[UNK]',
bos_id=2,
bos_piece='[CLS]',
eos_id=3,
eos_piece='[SEP]',
train_extremely_large_corpus=True
)
Training procedure
Models are pre-trained by UER-py on Tencent Cloud. We pre-train 1,000,000 steps with a sequence length of 128 and then pre-train 250,000 additional steps with a sequence length of 512. We use the same hyper-parameters on different model sizes.
Taking the case of word-based RoBERTa-Medium
Stage1:
python3 preprocess.py --corpus_path corpora/cluecorpussmall.txt \
--spm_model_path models/cluecorpussmall_spm.model \
--dataset_path cluecorpussmall_word_seq128_dataset.pt \
--processes_num 32 --seq_length 128 \
--dynamic_masking --target mlm
python3 pretrain.py --dataset_path cluecorpussmall_word_seq128_dataset.pt \
--spm_model_path models/cluecorpussmall_spm.model \
--config_path models/bert/medium_config.json \
--output_model_path models/cluecorpussmall_word_roberta_medium_seq128_model.bin \
--world_size 8 --gpu_ranks 0 1 2 3 4 5 6 7 \
--total_steps 1000000 --save_checkpoint_steps 100000 --report_steps 50000 \
--learning_rate 1e-4 --batch_size 64 \
--embedding word_pos_seg --encoder transformer --mask fully_visible --target mlm --tie_weights
Stage2:
python3 preprocess.py --corpus_path corpora/cluecorpussmall.txt \
--spm_model_path models/cluecorpussmall_spm.model \
--dataset_path cluecorpussmall_word_seq512_dataset.pt \
--processes_num 32 --seq_length 512 \
--dynamic_masking --target mlm
python3 pretrain.py --dataset_path cluecorpussmall_word_seq512_dataset.pt \
--pretrained_model_path models/cluecorpussmall_word_roberta_medium_seq128_model.bin-1000000 \
--spm_model_path models/cluecorpussmall_spm.model \
--config_path models/bert/medium_config.json \
--output_model_path models/cluecorpussmall_word_roberta_medium_seq512_model.bin \
--world_size 8 --gpu_ranks 0 1 2 3 4 5 6 7 \
--total_steps 250000 --save_checkpoint_steps 50000 --report_steps 10000 \
--learning_rate 5e-5 --batch_size 16 \
--embedding word_pos_seg --encoder transformer --mask fully_visible --target mlm --tie_weights
Finally, we convert the pre-trained model into Huggingface's format:
python3 scripts/convert_bert_from_uer_to_huggingface.py --input_model_path models/cluecorpussmall_word_roberta_medium_seq128_model.bin-250000 \
--output_model_path pytorch_model.bin \
--layers_num 12 --target mlm
BibTeX entry and citation info
@article{zhao2019uer,
title={UER: An Open-Source Toolkit for Pre-training Models},
author={Zhao, Zhe and Chen, Hui and Zhang, Jinbin and Zhao, Xin and Liu, Tao and Lu, Wei and Chen, Xi and Deng, Haotang and Ju, Qi and Du, Xiaoyong},
journal={EMNLP-IJCNLP 2019},
pages={241},
year={2019}
}