`bert_pretrain_extractor`

Module Contents

`BertPretrainFeatureExtractor`	Extract [CLS] embedding feature from any untrained bert-like models
`ManyBertPretrainFeatureExtractor`

class bert_pretrain_extractor.BertPretrainFeatureExtractor(model_name: str, max_length: int = 512, batch_size=64)

Bases: src.feature_extractors.base_extractor.BaseExtractor

Extract [CLS] embedding feature from any untrained bert-like models

generate_features(data: pandas.Series) → pandas.DataFrame

Generates features in batch-mode, obtained from untrained bert model.

Parameters: data – Series with full_text column
Returns: Dataframe, that have index - id’s from data, and columns - bert features

class bert_pretrain_extractor.ManyBertPretrainFeatureExtractor(model_names: List[str], max_length: int = 512, batch_size=64)

Bases: src.feature_extractors.base_extractor.BaseExtractor

generate_features(X: pandas.Series) → pandas.DataFrame

Generates features for model