incerto.data.create_balanced_dataloader

incerto.data.create_balanced_dataloader#

incerto.data.create_balanced_dataloader(dataset, batch_size=128, num_workers=4, pin_memory=True)[source]#

Create data loader with balanced class sampling.

Ensures equal representation of all classes in each batch.

Parameters:

  • dataset (Dataset) – Dataset with ‘targets’ or ‘labels’ attribute

  • batch_size (int) – Batch size

  • num_workers (int) – Number of workers

  • pin_memory (bool) – Whether to pin memory

Return type:

DataLoader

Returns:

DataLoader with balanced sampling

Contents