run_mlm module

Fine-tuning the library models for masked language modeling (BERT, ALBERT, RoBERTa…) on a text file or a dataset.

Here is the full list of checkpoints on the hub that can be fine-tuned by this script: https://huggingface.co/models?filter=fill-mask

class run_mlm.DataTrainingArguments(dataset_name: Optional[str] = None, dataset_config_name: Optional[str] = None, train_file: Optional[str] = None, validation_file: Optional[str] = None, test_file: Optional[str] = None, max_train_samples: Optional[int] = None, max_eval_samples: Optional[int] = None, max_seq_length: Optional[int] = 512, line_by_line: bool = False, pad_to_max_length: bool = False, overwrite_cache: bool = False, validation_split_percentage: Optional[int] = 5, preprocessing_num_workers: Optional[int] = None, keep_linebreaks: bool = False, mlm_probability: float = 0.15)

Bases: object

Arguments pertaining to what data we are going to input our model for training and eval.

dataset_config_name: Optional[str] = None

dataset_name: Optional[str] = None

keep_linebreaks: bool = False

line_by_line: bool = False

max_eval_samples: Optional[int] = None

max_seq_length: Optional[int] = 512

max_train_samples: Optional[int] = None

mlm_probability: float = 0.15

overwrite_cache: bool = False

pad_to_max_length: bool = False

preprocessing_num_workers: Optional[int] = None

test_file: Optional[str] = None

train_file: Optional[str] = None

validation_file: Optional[str] = None

validation_split_percentage: Optional[int] = 5

class run_mlm.ModelArguments(model_name_or_path: Optional[str] = None, model_type: Optional[str] = None, config_overrides: Optional[str] = None, config_name: Optional[str] = None, tokenizer_name: Optional[str] = None, cache_dir: Optional[str] = None, use_fast_tokenizer: bool = True, model_revision: str = 'main', use_auth_token: bool = False, freeze_token_embed: bool = False, pretrained_token_embed: Optional[str] = None)

Bases: object

Arguments pertaining to which model/config/tokenizer we are going to fine-tune, or train from scratch.

cache_dir: Optional[str] = None

config_name: Optional[str] = None

config_overrides: Optional[str] = None

freeze_token_embed: bool = False

model_name_or_path: Optional[str] = None

model_revision: str = 'main'

model_type: Optional[str] = None

pretrained_token_embed: Optional[str] = None

tokenizer_name: Optional[str] = None

use_auth_token: bool = False

use_fast_tokenizer: bool = True

run_mlm.main()

run_mlm.read_txt_embeddings(file_name)

Load pre-trained word embeddings :param filename: the path to the pre-trained word embedding file in Glove format :type filename: str

Returns

• wordEmbedding (numpy nd-array) – Numpy array of embeddings

• dictionary (dict) – Dictionary of word to index mappings