Movatterモバイル変換


[0]ホーム

URL:


Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Sign up
Appearance settings

Reproduced package based on Masked Language Model Scoring (ACL2020).

License

NotificationsYou must be signed in to change notification settings

Ryutaro-A/mlm-scoring-transformers

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

8 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

日本語 README

This package is a reproduced implementation ofMasked Language Model Scoring (ACL2020).

The original implementation uses the mxnet library, which does not support Japanese.

Therefore, we are releasing a version that can be used with the Masked Model published on Hugging Face.

We have not tried it on all models, but we believe that most of the pre-trained models can be used.

Installation

git clone https://github.com/Ryutaro-A/mlm-scoring-transformers.gitcd mlm-scoring-transformerspip install .

Get Started

  • To calculate scores for Japanese sentences.
importmlmtpretrained_model_name='cl-tohoku/bert-base-japanese-whole-word-masking'scorer=mlmt.MLMScorer(pretrained_model_name,use_cuda=False)japanese_sample_sentences= ['お母さんが行けるなら、わたしは行くのをやめるよ。うちから二人も出ることはないから。','お母さんが行けると、わたしは行くのをやめるよ。うちから二人も出ることはないから。','お母さんが行けたなら、わたしは行くのをやめるよ。うちから二人も出ることはないから。','お母さんが行けるのだったら、わたしは行くのをやめるよ。うちから二人も出ることはないから。','日本酒を飲めば、駅の反対側にある「××酒蔵」が一番だね。とにかく品揃えが抜群だよ。','日本酒を飲むなら、駅の反対側にある「××酒蔵」が一番だね。とにかく品揃えが抜群だよ。','日本酒を飲むんだったら、駅の反対側にある「××酒蔵」が一番だね。とにかく品揃えが抜群だよ。','日本酒を飲むと、駅の反対側にある「××酒蔵」が一番だね。とにかく品揃えが抜群だよ。',]scores=scorer.score_sentences(japanese_sample_sentences)print('input_sentence, score')forsentence,scoreinzip(japanese_sample_sentences,scores):print(sentence,score)# >> input_sentence, score# お母さんが行けるなら、わたしは行くのをやめるよ。うちから二人も出ることはないから。 -72.90809887713657# お母さんが行けると、わたしは行くのをやめるよ。うちから二人も出ることはないから。 -75.87569694537336# お母さんが行けたなら、わたしは行くのをやめるよ。うちから二人も出ることはないから。 -65.31722020490005# お母さんが行けるのだったら、わたしは行くのをやめるよ。うちから二人も出ることはないから。 -86.46473170552028# 日本酒を飲めば、駅の反対側にある「××酒蔵」が一番だね。とにかく品揃えが抜群だよ。 -85.50868926288888# 日本酒を飲むなら、駅の反対側にある「××酒蔵」が一番だね。とにかく品揃えが抜群だよ。 -81.26314979794296# 日本酒を飲むんだったら、駅の反対側にある「××酒蔵」が一番だね。とにかく品揃えが抜群だよ。 -82.7387441759266# 日本酒を飲むと、駅の反対側にある「××酒蔵」が一番だね。とにかく品揃えが抜群だよ。 -92.14111483963103
  • To calculate scores for English sentences.
importmlmtpretrained_model_name='bert-base-uncased'scorer=mlmt.MLMScorer(pretrained_model_name,use_cuda=False)english_sample_sentences= ['Due to the rain, our performance in the game was far from perfect.','Due to the rain, our performance in the game was apart from perfect.','Due to the rain, our performance in the game was different from perfect.','Due to the rain, our performance in the game was free from perfect.',]scores=scorer.score_sentences(english_sample_sentences)print('input_sentence, score')forsentence,scoreinzip(english_sample_sentences,scores):print(sentence,score)# >> input_sentence, score# Due to the rain, our performance in the game was far from perfect. -13.874692459549525# Due to the rain, our performance in the game was apart from perfect. -15.486674794020251# Due to the rain, our performance in the game was different from perfect. -16.62563831794064# Due to the rain, our performance in the game was free from perfect. -20.5683701854279

To change model config

Basically, the config used to pre-train the model is automatically selected, but you can also use your own config.

In that case, setmodel_config as follows.

config=transformers.BertConfig(hidden_act="gelu",hidden_size=1024,initializer_range=0.02,intermediate_size=4096,layer_norm_eps=1e-12,max_position_embeddings=512,model_type="bert",num_attention_heads=16,num_hidden_layers=24,pad_token_id=0,tokenizer_class="BertJapaneseTokenizer",type_vocab_size=2,vocab_size=32768,hidden_dropout_prob=0.2,attention_probs_dropout_prob=0.37)scorer=mlmt.MLMScorer(pretrained_model_name,model_config=config,use_cuda=False)

Options

2022/12/14 Added option to get per-token log-likelihood as wellSettingget_token_likelihood toTrue returns the total score and the per-token score as a dictionary.

scores=my_scorer.score_sentences(sentences=en,get_token_likelihood=True)print('input_sentence, score')forsentence,scoreinzip(en,scores):print(sentence,score["all"])print(score["token"])# Due to the rain, our performance in the game was far from perfect. -13.874687737519245# [-0.00044868520073119083, -0.0002509074244949909, -7.234254390419689, -0.1027699065355511, -0.05655604143014172, -0.04961800099545115, -0.0015554001203739796, -0.004590661092892022, -6.211619135159143, -0.21036846650855923, -0.0017955319970342492, -5.960464655174753e-08, -0.00011099000773481521, -0.00026807801725587353, -0.00048148300554723856]

License

This software is released under the MIT License, see LICENSE.txt.

Contacts

Twitter:@ryu1104_m

Mail: ryu1104.as[at]gmail.com

About

Reproduced package based on Masked Language Model Scoring (ACL2020).

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages


[8]ページ先頭

©2009-2025 Movatter.jp