pyproteonet.masking.masked_dataset_generator.MaskedDatasetGenerator

class pyproteonet.masking.masked_dataset_generator.MaskedDatasetGenerator(datasets: List[Dataset], generator_fn: Callable[[Dataset], MaskedDataset], sample_wise: bool = False, epoch_size_multiplier: int = 1, shuffle_samplewise_samples: bool = False)

Itarable set of MaskedDatasets and sample names. The MaskedDatasets are generated from a given list of Datasets using a given generator function that takes a Dataset and returns a MaskedDataset. The sample_names given with each MaskedDataset are either a list of all samples or just a single simple if sample_wise is True. This can be used by a DataLoader to generate MaskedDatasets on the fly.

Parameters:

  • datasets (List[Dataset]) – List of datasets to generate masked datasets from.

  • generator_fn (Callable[[Dataset], MaskedDataset]) – Function that generates a masked dataset from a given dataset.

  • sample_wise (bool, optional) – If True, returns each MaskedDataset num_samples times each times accompied by another sample name. If false just returns the MaskedDataset together with a list of all sample names. Defaults to False.

  • epoch_size_multiplier (int, optional) – Multiplier for the number of iterations over the datasets to make up one “epoch”. Defaults to 1.

  • shuffle_samplewise_samples (bool, optional) – If True and sample_wise is True, shuffles the samples within each dataset. Defaults to False.

__init__(datasets: List[Dataset], generator_fn: Callable[[Dataset], MaskedDataset], sample_wise: bool = False, epoch_size_multiplier: int = 1, shuffle_samplewise_samples: bool = False)

Methods

__init__(datasets, generator_fn[, ...])