# Copyright (c) Ye Liu. Licensed under the MIT License.
import torch.utils.data.dataset as dataset
from nncore import Registry, build_object
DATASETS = Registry('dataset')
[docs]
def build_dataset(cfg, *args, **kwargs):
"""
Build a dataset from a dict. This method searches for datasets in
:obj:`DATASETS` first, and then fall back to
:obj:`torch.utils.data.dataset`.
Args:
cfg (dict): The config of the dataset.
Returns:
:obj:`nn.Module`: The constructed dataset.
"""
return build_object(cfg, [DATASETS, dataset], args=args, **kwargs)