aiaccel.torch.datasets.CachedDataset¶

class aiaccel.torch.datasets.CachedDataset(dataset: Dataset[T_co])[source]¶

A dataset wrapper that caches the samples to improve performance.

Parameters:: dataset (Dataset) – The original dataset to be wrapped.

dataset¶

The original dataset.

Type:: Dataset

manager¶

The multiprocessing manager.

Type:: Manager

cache¶

The cache dictionary to store the cached samples.

Type:: dict

__init__(dataset: Dataset[T_co]) → None[source]¶

Methods

__init__(dataset)