mipcandy.data.dataset

Module Contents

Classes

KFPicker
OrderedKFPicker
RandomKFPicker
Loader
TensorLoader
_AbstractDataset
UnsupervisedDataset	Do not use this as a generic class. Only parameterize it if you are inheriting from it.
SupervisedDataset	Do not use this as a generic class. Only parameterize it if you are inheriting from it.
DatasetFromMemory
MergedDataset
ComposeDataset
PathBasedUnsupervisedDataset
SimpleDataset
PathBasedSupervisedDataset
NNUNetDataset
BinarizedDataset

Data

T
D

API

class mipcandy.data.dataset.KFPicker

Bases: object

abstractmethod static pick(n: int, fold: Literal[0, 1, 2, 3, 4, all]) → tuple[int, ...]

class mipcandy.data.dataset.OrderedKFPicker

Bases: mipcandy.data.dataset.KFPicker

static pick(n: int, fold: Literal[0, 1, 2, 3, 4, all]) → tuple[int, ...]

class mipcandy.data.dataset.RandomKFPicker

Bases: mipcandy.data.dataset.OrderedKFPicker

static pick(n: int, fold: Literal[0, 1, 2, 3, 4, all]) → tuple[int, ...]

class mipcandy.data.dataset.Loader

Bases: object

static do_load(path: str | os.PathLike[str], *, is_label: bool = False, device: mipcandy.types.Device = 'cpu', **kwargs) → torch.Tensor

class mipcandy.data.dataset.TensorLoader

Bases: mipcandy.data.dataset.Loader

static do_load(path: str | os.PathLike[str], *, is_label: bool = False, device: mipcandy.types.Device = 'cpu', **kwargs) → torch.Tensor

mipcandy.data.dataset.T = 'TypeVar(...)'

class mipcandy.data.dataset._AbstractDataset(device: mipcandy.types.Device)

Bases: torch.utils.data.Dataset, mipcandy.data.dataset.Loader, mipcandy.layer.HasDevice, typing.Generic[mipcandy.data.dataset.T], typing.Sequence[mipcandy.data.dataset.T]

abstractmethod load(idx: int) → mipcandy.data.dataset.T

Do not use this directly.

__getitem__(idx: int) → mipcandy.data.dataset.T

mipcandy.data.dataset.D = 'TypeVar(...)'

class mipcandy.data.dataset.UnsupervisedDataset(images: mipcandy.data.dataset.D, *, transform: mipcandy.types.Transform | None = None, device: mipcandy.types.Device = 'cpu')

Bases: mipcandy.data.dataset._AbstractDataset[torch.Tensor], typing.Generic[mipcandy.data.dataset.D]

Do not use this as a generic class. Only parameterize it if you are inheriting from it.

Initialization

__len__() → int

__getitem__(idx: int) → torch.Tensor

transform() → mipcandy.types.Transform | None

set_transform(transform: mipcandy.types.Transform | None) → None

class mipcandy.data.dataset.SupervisedDataset(images: mipcandy.data.dataset.D, labels: mipcandy.data.dataset.D, *, transform: mipcandy.data.transform.JointTransform | None = None, device: mipcandy.types.Device = 'cpu')

Bases: mipcandy.data.dataset._AbstractDataset[tuple[torch.Tensor, torch.Tensor]], typing.Generic[mipcandy.data.dataset.D]

Do not use this as a generic class. Only parameterize it if you are inheriting from it.

Initialization

_nd() → int

__len__() → int

abstractmethod load_image(idx: int) → torch.Tensor

abstractmethod load_label(idx: int) → torch.Tensor

load(idx: int) → tuple[torch.Tensor, torch.Tensor]

__getitem__(idx: int) → tuple[torch.Tensor, torch.Tensor]

image(idx: int) → torch.Tensor

label(idx: int) → torch.Tensor

transform() → mipcandy.data.transform.JointTransform | None

set_transform(transform: mipcandy.data.transform.JointTransform | None) → None

abstractmethod construct_new(images: list[Any], labels: list[Any]) → Self

preload(output_folder: str | os.PathLike[str], *, do_transform: bool = False) → None

fold(*, fold: Literal[0, 1, 2, 3, 4, all] = 'all', picker: type[mipcandy.data.dataset.KFPicker] = OrderedKFPicker) → tuple[Self, Self]

class mipcandy.data.dataset.DatasetFromMemory(images: Sequence[torch.Tensor], *, transform: mipcandy.types.Transform | None = None, device: mipcandy.types.Device = 'cpu')

Bases: mipcandy.data.dataset.UnsupervisedDataset[typing.Sequence[torch.Tensor]]

Initialization

load(idx: int) → torch.Tensor

class mipcandy.data.dataset.MergedDataset(images: mipcandy.data.dataset.UnsupervisedDataset, labels: mipcandy.data.dataset.UnsupervisedDataset, *, transform: mipcandy.data.transform.JointTransform | None = None, device: mipcandy.types.Device = 'cpu')

Bases: mipcandy.data.dataset.SupervisedDataset[mipcandy.data.dataset.UnsupervisedDataset]

load_image(idx: int) → torch.Tensor

load_label(idx: int) → torch.Tensor

construct_new(images: list[Any], labels: list[Any]) → Self

class mipcandy.data.dataset.ComposeDataset(bases: Sequence[mipcandy.data.dataset.SupervisedDataset] | Sequence[mipcandy.data.dataset.UnsupervisedDataset], *, device: mipcandy.types.Device = 'cpu')

Bases: mipcandy.data.dataset._AbstractDataset[tuple[torch.Tensor, torch.Tensor]| torch.Tensor]

Initialization

load(idx: int) → tuple[torch.Tensor, torch.Tensor] | torch.Tensor

__len__() → int

class mipcandy.data.dataset.PathBasedUnsupervisedDataset(images: mipcandy.data.dataset.D, *, transform: mipcandy.types.Transform | None = None, device: mipcandy.types.Device = 'cpu')

Bases: mipcandy.data.dataset.UnsupervisedDataset[list[str]]

paths() → list[str]

save_paths(to: str | os.PathLike[str]) → None

class mipcandy.data.dataset.SimpleDataset(folder: str | os.PathLike[str], is_label: bool, *, transform: mipcandy.types.Transform | None = None, device: mipcandy.types.Device = 'cpu')

Bases: mipcandy.data.dataset.PathBasedUnsupervisedDataset

load(idx: int) → torch.Tensor

class mipcandy.data.dataset.PathBasedSupervisedDataset(images: mipcandy.data.dataset.D, labels: mipcandy.data.dataset.D, *, transform: mipcandy.data.transform.JointTransform | None = None, device: mipcandy.types.Device = 'cpu')

Bases: mipcandy.data.dataset.SupervisedDataset[list[str]]

paths() → list[tuple[str, str]]

save_paths(to: str | os.PathLike[str]) → None

class mipcandy.data.dataset.NNUNetDataset(folder: str | os.PathLike[str], *, split: str | Literal[Tr, Ts] = 'Tr', prefix: str = '', align_spacing: bool = False, transform: mipcandy.data.transform.JointTransform | None = None, device: mipcandy.types.Device = 'cpu')

Bases: mipcandy.data.dataset.PathBasedSupervisedDataset

folder() → str

static _create_subset(folder: str) → None

load_image(idx: int) → torch.Tensor

load_label(idx: int) → torch.Tensor

save(split: str | Literal[Tr, Ts], *, target_folder: str | os.PathLike[str] | None = None) → None

construct_new(images: list[Any], labels: list[Any]) → Self

class mipcandy.data.dataset.BinarizedDataset(base: mipcandy.data.dataset.SupervisedDataset, positive_ids: tuple[int, ...], *, transform: mipcandy.data.transform.JointTransform | None = None, device: mipcandy.types.Device = 'cpu')

Bases: mipcandy.data.dataset.SupervisedDataset[tuple[None]]

__len__() → int

abstractmethod construct_new(images: list[Any], labels: list[Any]) → Self

load_image(idx: int) → torch.Tensor

load_label(idx: int) → torch.Tensor

fold(*, fold: Literal[0, 1, 2, 3, 4, all] = 'all', picker: type[mipcandy.data.dataset.KFPicker] = OrderedKFPicker) → tuple[Self, Self]