neuralset.dataloader.Batch

class neuralset.dataloader.Batch(*, segments: Sequence[Segment], data: dict[str, Tensor])

A collection of extracted features for a list of segments.

Returned by the SegmentDataset or DataLoader, this object holds both the underlying segments and the computed tensor data ready for machine learning models.

data

The extracted feature tensors. Keys match the names of the extractors passed to the Segmenter. Tensors are always batched along the first dimension.

Type:

dict of str to torch.Tensor

segments

The Segment instances corresponding to this batch. Useful for debugging or custom logic.

Type:

list of Segment

to(device: str) → Batch

Creates a new instance on the appropriate device