nessie.detectors.curriculum_spotter

Module Contents

Classes

CurriculumSpotter	Spotting Spurious Data with Neural Networks
CurriculumSpotterDataset
CurriculumSpotterDatasetCallback	Callback that is called during certain events (starting/ending training, ...)
CurriculumSpotterTransformerTextClassifier

class nessie.detectors.curriculum_spotter.CurriculumSpotter(verbose: bool = True, max_epochs: int = 24)

Bases: nessie.detectors.error_detector.Detector

Spotting Spurious Data with Neural Networks

Hadi Amiri, Timothy A. Miller, Guergana Savova https://aclanthology.org/N18-1182.pdf

error_detector_kind(self) → nessie.detectors.error_detector.DetectorKind

score(self, texts: nessie.types.StringArray, labels: nessie.types.StringArray, **kwargs) → numpy.typing.NDArray[float]

Apply curriculum learning and score the items by the perceived difficulty during the curriculum training.

Parameters

• texts – a (num_instances, ) string sequence containing the texts for each instance

• labels – a (num_instances, ) string sequence containing the noisy label for each instance

Returns

a (num_instances,) numpy array of bools containing the flags after using CS

Return type

scores

class nessie.detectors.curriculum_spotter.CurriculumSpotterDataset(tokenized_texts: Dict, encoded_labels: List[int])

Bases: torch.utils.data.Dataset

property X(self)

__getitem__(self, idx: int)

__len__(self)

property true_len(self)

update_mapping(self, new_dataset_mask: numpy.ndarray)

property y(self)

class nessie.detectors.curriculum_spotter.CurriculumSpotterDatasetCallback(model: nessie.models.text.transformer_text_classifier.TransformerTextClassifier)

Bases: transformers.TrainerCallback

Callback that is called during certain events (starting/ending training, …)

on_epoch_begin(self, args: transformers.TrainingArguments, state: transformers.TrainerState, control: transformers.TrainerControl, **kwargs)

on_epoch_end(self, args: transformers.TrainingArguments, state: transformers.TrainerState, control: transformers.TrainerControl, **kwargs)

on_train_begin(self, args: transformers.TrainingArguments, state: transformers.TrainerState, control: transformers.TrainerControl, **kwargs)

on_train_end(self, args: transformers.TrainingArguments, state: transformers.TrainerState, control: transformers.TrainerControl, **kwargs)

class nessie.detectors.curriculum_spotter.CurriculumSpotterTransformerTextClassifier

Bases: nessie.models.text.transformer_text_classifier.TransformerTextClassifier