RangeBatchSampler

class lightkit.data.RangeBatchSampler(sampler, batch_size, drop_last=False)[source]

Bases: torch.utils.data.sampler.Sampler[range]

Sampler providing batches of contiguous indices.

This sampler can be used with lightkit.data.DataLoader to provide significant speedups for tensor datasets.

Parameters

  • sampler (Sampler[int]) -- The sampler providing indices. Must be a sequential sampler. Note that the only purpose of this sampler is to determine its length.

  • batch_size (int) -- The number of items to sample for each batch.

  • drop_last (bool) -- Whether to drop the last batch if num_items is not divisible by batch_size.

Methods

Inherited Methods