basecls.data#

class basecls.data.FakeData[源代码]#

基类：object

Fake data useful for benchmark.

classmethod build(cfg, train=True, augments=None)[源代码]#
返回类型

DataLoader

class basecls.data.FolderLoader[源代码]#

基类：object

Local dataloader factory.

The source is the local folder.

classmethod build(cfg, train=True, augments=None)[源代码]#
返回类型

DataLoader

basecls.data.build_dataloader(cfg, train, augments=None, mode='folder', infinite=False, rank=None)[源代码]#

Build function for MegEngine dataloader.

参数

  • cfg (ConfigDict) – config for building dataloader.

  • train (bool) – train set or test set.

  • augments (Optional[Transform]) – augments for building dataloder. Default: None

  • infinite (bool) – make dataloader infinite or not. default: False

  • rank (Optional[int]) – machine rank, only useful for infinite dataloader. Default: None

返回类型

DataLoader

返回

A dataloader.

basecls.data.build_dataset(cfg, train=True, mode='folder')[源代码]#

Build function for dataset.

参数

  • cfg (ConfigDict) – config for building dataset.

  • train (bool) – train set or test set. Default: True

返回类型

VisionDataset

返回

A dataset which loads data from Nori on OSS.

class basecls.data.AutoAugment[源代码]#

基类：object

AutoAugment.

classmethod build(cfg)[源代码]#
返回类型

Transform

class basecls.data.ColorAugment[源代码]#

基类：object

Color augmentation.

classmethod build(cfg)[源代码]#
返回类型

Transform

class basecls.data.RandAugment[源代码]#

基类：object

Random augmentation.

classmethod build(cfg)[源代码]#
返回类型

Transform

class basecls.data.SimpleAugment[源代码]#

基类：object

Simple augmentation.

classmethod build(cfg)[源代码]#
返回类型

Transform

basecls.data.build_mixup(cfg, train=True)[源代码]#

Build (optionally) Mixup/CutMix augment.

参数

  • cfg (ConfigDict) – config for building Mixup/CutMix collator.

  • train (bool) – train set or test set. Default: True

返回类型

Optional[Collator]

返回

MixupCutmixCollator or None

basecls.data.build_transform(cfg, train=True, augments=None)[源代码]#

Build function for MegEngine transform.

参数

  • cfg (ConfigDict) – config for building transform.

  • train (bool) – train set or test set. Default: True

  • augments (Optional[Transform]) – augments for building transform.

返回类型

Transform

返回

A transform.