nessie.models.tagging.transformer_sequence_tagger

Module Contents

Classes

SequenceTaggingEvalDataset
SequenceTaggingTrainDataset
TransformerSequenceTagger	Helper class that provides a standard way to create an ABC using

class nessie.models.tagging.transformer_sequence_tagger.SequenceTaggingEvalDataset(tokenized_texts)

Bases: torch.utils.data.Dataset

__getitem__(self, idx)

__len__(self)

class nessie.models.tagging.transformer_sequence_tagger.SequenceTaggingTrainDataset(tokenized_texts, encoded_labels)

Bases: torch.utils.data.Dataset

__getitem__(self, idx)

__len__(self)

class nessie.models.tagging.transformer_sequence_tagger.TransformerSequenceTagger(verbose: bool = True, max_epochs: int = 24, batch_size: int = 32, model_name: str = BERT_BASE)

Bases: nessie.models.model.SequenceTagger, nessie.models.model.Callbackable

Helper class that provides a standard way to create an ABC using inheritance.

add_callback(self, name: str, callback: transformers.TrainerCallback)

fit(self, X: nessie.types.RaggedStringArray, y: nessie.types.RaggedStringArray)

has_dropout(self) → bool

label_encoder(self) → sklearn.preprocessing.LabelEncoder

Returns a label encoder that can be used to map labels to ints and vice versa

predict(self, X: nessie.types.RaggedStringArray) → awkward.Array

predict_proba(self, X: nessie.types.RaggedStringArray) → awkward.Array

Returns a distribution over labels for each instance.

Parameters

X – The token sequences to predict on

Returns

A (num_sentences, num_tokens, num_labels) ragged array

score(self, X: nessie.types.RaggedStringArray) → awkward.Array

Returns the best score for each item

use_dropout(self, is_activated: bool)